A car is a difficult thing to steal. It is large, cherished by its owner, difficult to hide and required by law to have identifying metal plates that everyone can see. A good thief changes the plates, and an excellent thief steals a common car in a boring color and hopes to blend in among the crowd. But eventually, no matter what techniques they use, they must drive that car on the road, and roads are public places.

But until recently, finding a stolen car and catching its thief depended on a diligent police officer looking at the right road at the right moment and managing to copy down a license plate number going past at speed, and then remembering he had seen it on the morning briefing’s stolen car hot sheet. America has four million miles of public roads, and almost 50 percent of the country’s police departments employ fewer than ten full-time officers. The odds were in the thief’s favor.

With not much to go on, police officers were forced to operate on vague descriptions and partial plates. While I pulled over a car that happened to be the same color as the suspect’s vehicle, and spent time checking names and licenses; the criminal was usually somewhere else. Imprecision allowed thieves to escape while intruding on the lives of millions of innocent motorists. That is, until the arrival of systems like Flock.

Bare ALPR

Flock Safety, a startup based in Atlanta, Georgia, was founded in 2017 to give police departments better eyes. Its product is a small solar-powered automatic license plate recognition (ALPR) camera attached to a pole on the roadside. As a car passes, the device takes a photograph of the vehicle, notes identifying features like color and make, and reads its license plate. The license plate is instantly checked against the FBI’s national database of stolen cars or wanted persons. If there’s a match, it sends an alert to police officers in the area, who may be eating their lunch instead of watching the road. As a police officer in the southeastern United States, I have often used Flock to catch wanted criminals.

But ALPR is not new. Originally developed in 1976 by the British Home Office, it was deployed as a prototype on toll roads and tunnels, leading to the first ALPR-enabled arrest in 1981 when a stolen car being driven by its thief was found. The first large-scale trial of the technology was ‘Project Laser’ in 2004, when the project’s cameras read more than five million British plates, resulting in 3,071 arrests. Just a year later, ALPR was used to locate a man who murdered a female police officer. At the time, Chief Superintendent Geoff Dodd of the West Yorkshire Police described it as ‘the best investigative tool we have had since the introduction of DNA analysis’.

Criminals have learned to fear the technology. In London, where I spent five years as a police officer before moving to America, ALPR is so thoroughly embedded in criminal life that rappers discuss it in their lyrics. ‘Bare A[L]PR, bro,’ the artist Central Cee raps, ‘beg you don’t smoke weed in the car’.

For much of the technology’s history, however, it has needed expensive installations or cameras attached to police vehicles, affordable to only the largest departments. Alerts were often neither instant nor accurate. Flock’s idea was to sell a more accurate ALPR camera for an annual subscription of just $3,000, an affordable price given that the typical US police department has an annual budget of $1 million. It worked, and today almost 30 percent of police agencies in the United States subscribe to the service.

It is easiest to imagine the Flock camera, or any ALPR device, as a roadside barcode scanner. I can use a ‘lookup’ tool for either a plate or vehicle description that could be connected to a specific investigation. For instance if a victim of a sexual assault told me the suspect was driving a white Toyota Tacoma with a roof rack, I could while still on scene conduct a lookup in the area filtered for similar vehicles with roof racks and see results from within a particular timeframe. For each search, I am required to record a case number and a reason that is then published in a monthly audit sent to and validated by department administrators.

I have used Flock myself, for instance, to locate a vehicle involved in a hit and run when all there was to go on was a model and color. Using that and an approximate time window, I was able to find an image from when a car entered my jurisdiction, and when it left, with the visible addition of a significant dent from the collision.

What Flock cannot do is search for individuals or show who is driving a particular vehicle that it scans. An ALPR camera does not care about the person driving a car, or its passengers, and even if an officer gets an alert that a vehicle is known to be driven by a wanted felon they must themselves establish who is behind the wheel before having probable cause to stop it.

Stolen cars provide the clearest evidence such a scanner works. American police solve only 8.2 percent of reported vehicle thefts with an arrest. A recent working paper suggests that agencies that adopted the devices saw a 15.9 percent relative increase in car thefts caught, which would raise the national rate to 9.5 percent. But that is just stolen cars.

Vehicle crime takes many forms. It is often the means by which a robber or a murderer arrives at and departs from their crimes. Criminals’ vehicles carry drugs and guns, and smuggle cash. In Atlantic City, New Jersey, vehicles were involved in 53 percent of shootings during the two years before the city expanded its ALPR network. After the expansion, Atlantic City saw monthly averages of motor vehicle thefts drop by 20 percent, property crimes by 34 percent and fatal shootings by almost 40 percent. Seventy-two alert-caused traffic stops located forty stolen vehicles and nine stolen plates.

Flock itself conducted a study claiming that 10 percent of all reported crime in the United States is solved using evidence obtained from their cameras. Not bad for a few thousand dollars a year.

On several occasions, Flock alerts have allowed me to arrest those suspected of serious violent crimes. Local officers knew who they were and had circulated information about what cars they drove to neighboring agencies. Besides automatically cross-referencing from national databases, officers can create their own ‘hotlists’ of license plates that will create alerts, even if they are not on FBI-maintained lists.

These could be hotlists to build intelligence on drug supply or gangs, or plates belonging to drivers in the city with suspended licenses. These officer-created hotlists can be shared locally or nationally, but still require an officer to verify the information and establish lawful reasons to investigate the cars involved. Without those alerts, some of the criminals I arrested would have remained free for longer.

Share

Good Flock, Bad Flock

And yet, the past month has seen this simple piece of crime-fighting technology become the most reviled piece of street furniture in America. Flock cameras have been sawn from their poles, hammered into shards by teenagers, and rammed by vehicles. Cities have canceled contracts even where their own audits found no evidence that their officers had misused the system. Opposition stretches from Bernie Sanders all the way to the former WWE wrestler Kane, who now serves as the Republican mayor of Knox County, Tennessee, and who describes Flock as ‘unconstitutional’.

More than 150 cities and towns have now deactivated their Flock cameras or canceled contracts with the company. Over the space of a few weeks a startup previously known to just police officers and neighborhood associations has joined data centers, Covid vaccines, 5G towers, pasteurized milk and fracking in the pantheon of American moral panics.

Opponents of Flock tend to believe that it is abused by policemen with impunity, that it can track individuals as well as cars, and that it violates American constitutional protections. None of those things are true.

Obviously a camera capable of finding a stolen car is also capable of finding a car driven by somebody’s ex-wife and so like any software, Flock has been abused. There are stories of police officers who have used it to stalk girlfriends, and there are also innocent motorists who have been stopped by police after Flock cameras misread plates or received old information from national databases. A common issue from my experience is stolen front license plates being entered into a database and the innocent owners of the rear license plate being held on suspicion of car theft.

But Flock comes with protections. As mentioned earlier, each search in Flock must be accompanied by a recorded reason. Similarly entering a plate into a hotlist must have a case number assigned. Results were originally retained for thirty days, but recent changes by Flock mean that recorded plates are now kept for only seven days. Suspicious searches are picked up by algorithms or found in department audits, with the service blocking users until senior officers evaluate and resolve flags. The Institute for Justice, a think tank that is critical of ALPR, studied misuse of Flock in April of this year and found 51 incidents across the United States since 2024, noting that ‘Nearly all of these officers were criminally charged and lost their jobs, either by resigning or getting fired’. Another incorrect belief is that Flock does more than scan vehicles, with some suggesting it scans and tracks passing phones or devices – but it does no such thing.

While Flock is a tool that can be audited, there is nothing to stop a corrupt officer simply following a car when they don’t like the look of its driver, or writing down plates of vehicles spotted outside an ex-girlfriend’s house – they would just be harder to catch. And Flock does not proactively alert officers to cars that do not trigger flags on national or local hotlists. There is no reason state legislatures or Congress could not create harsh punishment or penalties for abuse of ALPR, without getting rid of it entirely.

Another argument against Flock, as espoused by Kane, is that it violates liberties granted by the Fourth Amendment, protecting citizens from unreasonable privacy breaches from law enforcement. But it has been repeatedly established over the last century by courts across the country that cops observing license plates, either with the naked eye or by machine, is not an unreasonable search. After all, your license plate belongs to the government and a highway is a public place.

In 1983, the Supreme Court held unanimously that ‘a person travelling in an automobile on public thoroughfares has no reasonable expectation of privacy in his movements from one place to another’. And in January of 2026, a federal judge in Virginia ruled in a challenge by two civil rights activists against Flock specifically that: ‘An individual scan of a license plate, a vehicle identifier that generally must be displayed under state law, does not invade any privacy interests’.

The problem for Flock is that, if it works, it provides no visible benefits outside the crime statistics: a crime prevented because an offender was afraid of being caught by ALPR produces no victim and no police report. But every misuse has a victim who can describe what happened and a mugshot of a corrupt cop, and every false alert can produce a photograph of an innocent motorist in handcuffs.

Flock did not invent its capabilities and certainly does not have a monopoly on them. There are at least five other companies that offer almost identical products and services to public and private enterprise, some arguably even more invasive. Even some of the cities that have canceled contracts with Flock Safety in recent weeks have signed up to buy similar products from the company’s rivals.

It seems unlikely that America will ever ban ALPR at a state level, let alone a national one. Instead, a patchwork already familiar to American policing will result, where policies and practices diverge across local and political boundaries.

Nor would the cameras actually disappear, even in towns where governments restrict their use and cancel contracts. The already-mentioned Fourth Amendment only applies to the government, not to private enterprise. Police departments may dismantle their networks but homeowners associations concerned about vehicle theft can buy their own network of Flock cameras, as many already do. The people harmed the most will be the residents of the average small town whose police departments cannot afford helicopters, drones or enough police officers to watch every road themselves.

I can remember policing without ALPR. A circulation would go out over the radio to look for a silver Volkswagen Golf, no license plate seen, and so for the rest of the night every single silver Golf we saw got stopped. There was no audit trail of our mistakes because there was no software.

Even when ALPR was introduced, we relied on a complex system that would send alerts through multiple layers of human filtering before finally being sent out to officers – often five to ten minutes after the car had been detected – and so was usually long gone by the time we turned up. With Flock, I can get a notification on my phone and can set my patrol car up to intercept it as it approaches.

Flock’s critics have the important point right: we should want the police to invade law-abiding citizens’ lives as little as possible. But they are mistaken about what Flock actually does. Flock allows us to be more surgical and precise, avoiding pulling over innocent people with all of the risks that involves. More than anything, Flock allows me to leave you alone.

Ned Donovan is a policeman

Subscribe now

Jacob Witten explains why siRNA will change medicine as we know it.

If you want a picture of biomedical research, imagine a boot stamping on a biologist’s face – for decades. The boot is labeled ‘medicinal chemistry’. This should not be taken as an insult to medicinal chemists, who heroically navigate maybe the hardest field in biomedicine.1 Rather, what I mean is this. A biologist might have a clear hypothesis about how to stop a disease, such as: blocking a particular protein will lower blood cholesterol and reduce the risk of heart disease. Naturally enough, the biologist would like to test that hypothesis. But there is no magic button to switch off this protein; something physical has to do the blocking. Usually, that thing is a small chemical carefully designed to be administered into the body, bind to the protein, block it – and do nothing else. In practice, this almost never goes as planned, which is a major reason most drugs fail in clinical trials.

It’s worse than this: medicinal chemists struggle to come up with ways to interact with most proteins at all. The majority of human proteins are considered ‘undruggable’, meaning that none of the small molecules we can currently synthesize can effectively bind to them, either because those proteins lack a convenient pocket for drugs to attach to or because they are ‘intrinsically disordered’, with little coherent structure.2 Drug discovery ends up focusing on the minority of targets that are druggable, like the proverbial drunk searching for his keys under the streetlamp because that’s where the light shines.

Even for those targets, success is far from guaranteed. A drug that works in the lab might be broken down in the body before it reaches its target protein, interfere with other proteins and cause toxic side effects, or fail to be absorbed and distributed to the right tissues. Our ability to predict what small molecule drugs will do in the body is so poor that even if the drug works, it might be for a different reason than originally hypothesized!3

But what if, instead of spending years optimizing small molecule chemistry to treat just one disease, and all that work merely delivering a sharp reminder that no plan survives first contact with the enemy, there really was a magic button to switch off any gene we chose? In that case, we could treat not just high cholesterol but dozens of deadly diseases, from Alzheimer’s to diabetes to Huntington’s disease.

Welcome to the world of siRNA therapy.

Silencing genes

The story of siRNA begins in 1990, when the scientists at the DNA Plant Technology Corporation in California were trying to figure out why their petunias turned white. They had actually been trying to make their petunias darker, by adding an extra copy of the gene for an enzyme that produces pigment. Presumably, adding an extra copy of a gene for an enzyme would mean more enzyme and therefore more pigment. But somehow, the extra copy eliminated the enzyme’s production, rather than boosting it. They named this baffling effect ‘cosuppression’. Subsequent work by other researchers revealed its cause: a special kind of RNA molecule, siRNA, had interfered with the enzyme’s production. Andrew Fire and Craig Mello later won a Nobel prize for this discovery.

Biologists had already known about RNA for several decades. Similar to DNA, RNA is a complex molecule built from a chain of smaller building blocks that encode genetic information. But the two differ in one important detail: the building blocks of RNA have one extra oxygen atom compared to DNA (hence its name, ribonucleic acid, as opposed to DNA’s ‘deoxy’-ribonucleic acid). This oxygen is prone to chemical reactions that can break RNA molecules. DNA is stabler and is therefore used to store genetic information in all multicellular life. RNA molecules tend to be short-lived, and take many different forms within cells, carrying instructions, helping to assemble proteins, regulating genes, and catalyzing reactions.4 And while DNA is ‘double-stranded’, with two linked strands of building blocks twisting into its famous helix shape, RNA often exists as just one strand.

DNA contains the instructions for making proteins, from the enzymes that digest food to the keratin and collagen that form our hair and skin. But to get from DNA to protein, the genetic code is first transcribed into an intermediate RNA molecule, called ‘messenger RNA’ or mRNA, which is then translated into protein. If the mRNA is destroyed, the protein won’t be built.

That’s exactly what siRNA, or ‘small interfering RNA’, does: it’s a short strand of RNA that interferes with the production of protein by destroying mRNA. An siRNA molecule binds to a piece of mRNA with a matching sequence, like a kind of barcode, and targets it for destruction by the cell’s gene-silencing machinery, which cuts the mRNA into pieces.

siRNA also explains what happened in the petunias. Adding an extra copy of the pigment-producing gene triggered the petunias to develop siRNA targeting that sequence, and because the newly introduced gene and the original gene shared the same sequence, both were silenced, the flowers lost their pigment and turned white. This is a naturally occurring process: siRNA is used to ‘silence’ unwanted genes, such as those of viruses that have entered the cell. But we can also use it to artificially block the production of proteins that cause disease.

Subscribe now

siRNA therapy

For some diseases, gene silencing has already succeeded. One example is hereditary transthyretin amyloidosis, a condition caused by a mutation in the transthyretin gene, which makes that protein misfold and clump together in the bloodstream, nerves, and heart. Over the years, the deposits build up, damaging nerves and stiffening the heart muscles, which can eventually lead to heart failure. Until recently, the condition meant a life expectancy roughly a decade shorter than average.5

In 2018, the FDA approved the very first RNA interference drug, called patisiran, to treat the nerve damage caused by this disease. A newer, longer-lasting version called vutrisiran was approved in 2022. Both drugs use siRNA to prevent the production of the misfolded protein so that clumps don’t form, which halts or slows down progression of the disease.

This approach could be adapted to many different diseases, from high cholesterol to Alzheimer’s, whose progression hinges on a small number of critical proteins. RNA interference makes it possible to target those proteins specifically and stop them from being produced.

Have a gene you’d like to silence? Identify the gene’s sequence, synthesize a short siRNA sequence to match it, deliver the siRNA into the right cells, and bam – instant knockdown. It sounds simple enough in theory, but that last step, getting siRNA into the right cells, has proven surprisingly difficult. Only after decades of grinding organic chemistry has it been solved.

The delivery problem

If you just injected siRNA into a patient’s veins, nothing useful would happen. There are at least three reasons for this. First, the body is full of enzymes that recognize and destroy unprotected RNA, including siRNA. Second, even a perfectly intact siRNA molecule would do nothing in the bloodstream, as it needs to get inside a cell to reach its target mRNA. Third, the way it would get into a cell involves being enveloped in a small bubble called an endosome, where it would be broken down by digestive enzymes. So an siRNA molecule would need to stay intact, enter its target cells, and then somehow escape from the endosome without being digested.

One way to solve these challenges is to use nature’s own delivery system: viruses. Viruses have spent billions of years evolving to get past our cells’ defences, find their way in, and release their proteins. So what if we used viruses to carry siRNA into our cells? Researchers have tried to do just that, by engineering harmless viruses to carry and deliver siRNA into specific tissues, including the brain. But viruses come with a drawback: just as they have evolved to infiltrate our cells, our immune systems have spent hundreds of millions of years learning to recognize and destroy them. Even carefully designed viral delivery systems can trigger immune reactions, especially if they’re used repeatedly, which can be risky for patients and make treatments less effective.

Signed, sealed, delivered

Another approach has focused on packaging RNA inside tiny fat-like particles, called liposomes, to protect and help it enter cells. Liposomes were first explored in the 1960s, and could shield RNA from being broken down in the body to some degree. But liposomes were far less successful at entering cells, because they bound only weakly to their outer membranes, and were even worse at escaping endosomes and releasing their payload of siRNA into the cells.

That challenge was overcome in the 1980s, when researchers tweaked liposomes to bind more tightly to negatively charged cell membranes, by incorporating lipids with a positive electric charge. While this helped get the RNA into cells, it came at a cost: the strong positive charge made them interact with cell membranes too strongly and damage them.

The next breakthrough was more subtle. Instead of using lipids that were always charged, researchers incorporated lipids that could change their charge depending on the acidity of their environment, called ‘ionizable lipids’. These lipids are neutral in the bloodstream, minimizing damage, but turn positive once inside an endosome, as endosomes are acidic. This destabilizes the endosome’s membrane, releasing the siRNA into the cell.

Even after coming up with that basic idea, the problem was nowhere near solved. It took more than a decade of tinkering with ionizable lipid chemistry, leading to a thousand-fold improvement in potency, to produce lipid nanoparticles that performed all these three steps of delivery (stability, uptake into cells, and endosomal escape) safely and efficiently.6

Patisiran, the first FDA-approved siRNA drug, was the first success that took this approach, by delivering siRNA in ionizable lipid nanoparticles to silence the mutated transthyretin gene in liver cells and slow the progression of amyloidosis. It was proof that siRNA could be turned into a real medicine. But it also arrived, ironically, just as the use of lipid nanoparticles for siRNA was being eclipsed by a more efficient method.7 What if siRNA didn’t need a carrier at all?

Naked siRNA

So far, researchers had focused on shielding siRNA from being broken down inside the body. But then they tried a new tack: making the siRNA molecule itself sturdier so that it could survive unclothed.

To understand how this was possible, let’s return to the difference between DNA and RNA. RNA has an extra oxygen atom, which makes it vulnerable to reactions that break it apart. By chemically modifying that pesky oxygen – for example, by swapping the oxygen for less reactive fluorine or hydrogen, or linking a methyl group (a highly stable formulation of one carbon atom and three hydrogen atoms) to it – the RNA molecule can be made much more durable. These small, precise chemical changes allowed Alnylam Pharmaceuticals (which also developed patisiran) to create siRNA molecules with almost indefinite protection against degradation when injected naked.

With stability solved, the next challenge was guiding this naked siRNA to its target. The siRNA drugs carried by lipid nanoparticles naturally accumulated in the liver, because lipid nanoparticles impersonate cholesterols which get processed there. This made the liver an easy target for siRNA drugs. But without lipid nanoparticles, naked siRNA molecules wouldn’t concentrate at a target organ. So researchers attached a sort of address label to get them there: three copies of a small sugar molecule called GalNAc. Liver cells have receptors for GalNAc on their surfaces, which recognize the sugar and bring it inside their endosomes. This directed siRNA to its intended destination.

Alnylam had cracked the first two hurdles of delivering naked siRNA drugs: it had created siRNA that did not instantly break down in the bloodstream, got to the cells it needed to, and entered them. That left the final hurdle of escaping the endosome. Without ionizable lipids to break the endosome membrane, how would naked siRNA ever make it out of endosomes and into the interior of the cell? The answer turned out to be unexpectedly simple. While most naked siRNA remains confined in the endosome, a small proportion slips out through some fluke occurrence, and more continues to be released over time. The proportion of siRNA leaking out of endosomes can be well under one percent, but often that is enough because siRNA is so potent – very little of it is actually needed to have its gene-silencing effect.

The result is a massive difference in how long the drug lasts. Patisiran, which is carried by a lipid nanoparticle, needed to be dosed every three weeks. In contrast, Alnylam’s newer drug, vutrisiran, a naked siRNA drug attached to a GalNAc sugar, is approved to treat the same medical condition, but needs to be dosed only once every three months. Other drugs based on naked siRNA and GalNAc now exist that only need to be topped up every six months. GalNAc siRNA drugs also have another advantage. Patisiran, with its lipid nanoparticle, carried the risk of triggering immune reactions, so patients had to be given immunosuppressants alongside it; but this precaution is unnecessary for those attached to GalNAc.

Simply put, temporarily switching off a precise gene in the liver is now a solved problem, and it can be achieved with drugs that require no more than four administrations per year. Multiple siRNA drugs have been approved for treating liver diseases, both rare and common, including treatments for hereditary transthyretin amyloidosis, hemophilia, high cholesterol, and acute hepatic porphyria, a nasty lifelong condition that causes attacks of pain, vomiting, and breathlessness because of a faulty enzyme. Many more siRNA drugs are in late-stage trials, including treatments for high blood pressure, diabetes, and fatty liver disease.

Subscribe now

Beyond the liver

Every success story so far has involved the liver. But of course the liver is not the only organ in the body. The reason the liver dominates the story is not that siRNA will work only there but that the liver was an easy starting point. But once that approach worked, it suggested a broader idea – if siRNA can be stabilized and delivered by attaching a molecular ‘address’ to it, then in principle, different labels could be used to guide it to different organs.

That’s exactly what the field is now trying to do. By swapping out the sugar used to target the liver for other molecules, researchers are developing drugs targeted for other organs, including the kidneys, lungs, fat tissue, muscle, and brain. The competition is intense, and if any company stumbles, many other hungry startups, like Manifold Bio and Soufflé Therapeutics, will be hot on its tail.

Among these efforts, the highest-stakes trials are those aimed at the brain. They target one of modern medicine’s bêtes noires: Alzheimer’s disease. The most influential theory of Alzheimer’s, though it remains controversial, is that the disease is driven by a misfolded protein, amyloid beta, which is produced from a larger amyloid precursor protein. According to the theory, these misfolded proteins accumulate into clumps that trigger the accumulation of another protein, tau, which causes a loss of neurons and cognitive decline. This hypothesis naturally suggests two obvious intervention points: either reduce the production of amyloid precursor protein upstream or the production of tau downstream. Alnylam’s program targets the amyloid precursor, while Arrowhead’s targets tau.

These drugs will face different challenges than earlier siRNA drugs, which targeted the liver. The most immediate is accessing the brain, which is protected by the blood-brain barrier, a tight layer that seals and defends the brain from most molecules in the bloodstream. Alnylam’s current strategy is to bypass this barrier by injecting the siRNA directly into spinal fluid by ‘intrathecal injection’, where a needle inserted into the spine delivers drugs directly into spinal fluid, like in an epidural. Their siRNA is also attached to a fatty acid chain, which they believe will successfully deliver it to the relevant brain cells. But while this approach might prove effective, it’s also uncomfortable for patients and not ideal for frequent dosing.

Arrowhead is trying the bolder approach of delivering siRNA by injection under the skin, using a proprietary targeted siRNA molecule that’s meant to cross the blood-brain barrier itself. There’s a precedent for this approach, but no guarantee it will succeed.

Even if the amyloid hypothesis is correct and delivery succeeds, biology might still limit what siRNA can achieve. Famously, by the time Alzheimer’s disease is diagnosed, the damage to the brain is often extensive enough that clearing amyloid plaques has little effect on slowing its progress. Whether targeting tau might be more effective at later stages remains an open question. Yet the broader point still stands: siRNA delivery beyond the liver is a matter of when, not if. If biologists can identify proteins that are critical to the progression of a disease, those diseases can become treatable or preventable.

The future of programmable medicines

siRNA falls into the category of therapies we could call ‘programmable medicines’. Once medicinal chemists have worked out how to get a programmable medicine into a particular organ, all that’s required to treat a new disease of that organ is to specify a new RNA or DNA sequence while holding the chemistry constant.8 9 So vutrisiran, the siRNA treatment for hereditary transthyretin amyloidosis, works according to the exact same principles as inclisiran, the siRNA that targets PCSK9 to reduce cholesterol levels.

siRNA is not the only programmable medicine in use, and other platforms have capabilities that siRNA lacks. For example, antisense oligonucleotides (ASOs) are short strands of synthetic genetic material that can alter RNA and be delivered similarly to siRNA. One such ASO, targeting the tau protein involved in Alzheimer’s disease, has shown promising results in a phase two clinical trial. mRNA vaccines can trigger the production of certain proteins in our cells, and are a blindingly fast way to develop effective vaccines against viruses, as we saw during the Covid pandemic. CRISPR-based gene editing can permanently fix genes, rather than just temporarily switching them off, by directly altering cells’ DNA.10 The common thread is that all these approaches link biological understanding directly to medicine. Once delivery is solved, the medicinal chemistry boot is off the biologist’s face, and they can test new treatments simply by switching out the target genetic sequence.

Each of these forms of programmable medicine has already produced FDA-approved treatments. So far, only one class of problems has been solved by programmable medicine: gene silencing in the liver by siRNA. But sooner or later, the same will be true of other organs. As scientists solve each delivery problem, we’ll experience an explosion of safe, convenient treatments against a myriad of diseases that were once considered untreatable. Soon, they will simply be waiting their turn.

Jacob Witten is a professor at Case Western Reserve University.

Subscribe now

Derek Hopper explains why Thailand was one of the only countries not to be colonized.

At the beginning of the 1833 dry season, a monk named Mongkut set out for Sukhothai, the old capital of Siam. Long journeys by boat and on foot for study, teaching, and religious observance were common for monks like Mongkut. Theravada Buddhism was a huge part of life in premodern Siam, and at any given time, a small percentage of adult men were ordained as monks. As they traveled around the kingdom, locals provided them with food, clothes, and lodging.

But Mongkut was no ordinary monk. He was the 43rd child of King Rama II. Born to one of Rama II’s senior consorts, he was a high-ranking son and in the running to inherit the throne. As such, he would have been traveling with a retinue of attendants as well as fellow monks. The kingdom to which Mongkut was an heir covered most of modern-day Thailand, plus parts of Laos, Cambodia, and Malaysia, and contained around five million subjects. Though most of these subjects were poor rice farmers, Siam’s largest city, Bangkok, was a major trading port, with a dense network of canals lined with houses on stilts, and an ethnically diverse population.

In contrast to the bustling capital, Sukhothai was by this time largely abandoned. From the thirteenth to fifteenth centuries, Sukhothai had flourished as a political, cultural, and religious center, but it had since been eclipsed by Ayutthaya to the south and, since its founding in the eighteenth century, Bangkok. When Mongkut arrived there, he would have been greeted, as visitors are today, by crumbling stupas (religious buildings), serene Buddha statues, and rust-colored brickwork weathered by centuries of blistering sunshine and tropical rains.

At some point during his stay, Mongkut came across a remarkable object: a four-sided, black stone monolith, about a meter high and inscribed with an unusual script. The stone would become known as the Ramkhamhaeng Stele. Purported to date from 1292, it is often cited as the oldest surviving example of Thai writing.

Mongkut brought the Stele to Bangkok and assembled a committee of scholars to decipher its enigmatic inscription. They compared it to known forms of old Thai and other related scripts, analyzed linguistic patterns, and collaborated with learned monks to interpret unfamiliar words and characters. What they found was extraordinary: the 124 lines of archaic text told of an ancient Thai kingdom ruled by a wise monarch, Ramkhamhaeng. According to the Stele, Ramkhamhaeng’s kingdom combined a literary tradition, a sophisticated legal and political system, independence, sovereignty, secure geographic boundaries, Buddhist kingship, and just rule.

The discovery of the Stele was, as it turned out, serendipitous. Over five centuries, European powers colonized virtually every other territory on Earth, with exceptions that can be counted on the fingers of one hand. Japan is the best known example. Siam, perhaps the next clearest case of a state that resisted foreign occupation and retained sovereignty, is much less publicized. Siam held out through a combination of diplomacy, reform, and careful image management. By providing evidence of a longstanding complex civilization, the Ramkhamhaeng Stele was just the sort of brand asset the kingdom needed.

The Siamese Holdout

By the 1830s, the Siamese court was watching nervously as British power radiated outward from India. Just a few years before Mongkut’s discovery, during the First Burmese War of 1824–1826, British gunboats sailed up the sacred waters of the Irrawaddy River and humiliated the ancient kingdom of Burma, extracting huge territorial concessions from Siam’s long-term rival. To the east, French missionaries were active in Vietnam and Cambodia, representing the earliest phase of France’s colonial ambitions and its mission civilisatrice. One by one, Siam’s neighbors began to fall and it found itself the sole Southeast Asian holdout against European colonialism.

The vanquished countries pursued a range of resistance strategies, all of them unsuccessful. Burma fought a series of doomed wars against Britain, only to be annexed. Cambodia willingly allowed itself to become a French protectorate, viewing this as the lesser of several evils after decades of encroachment from neighboring Siam and Vietnam, only for that protection to evolve into colonial rule within French Indochina. Vietnam relied on the country’s historic ties to China to ward off France, but its harsh treatment of missionaries and minorities gave Europeans an easy pretext for intervention.

Instead, Siam’s leaders hoped that demonstrating it already had clear markers of civilization would undermine the pretexts for colonization. Even in the nineteenth century, countries did not want to be thought of as engaging in naked land grabs. Successful colonial ventures usually combined ostensibly noble justifications and straightforward power-hungry motivations. Europeans believed that technologically advanced, Christian nations had a moral duty to bring progress to ‘backward’ peoples who lacked proper government, writing systems, and cultural refinement. This ideology created domestic political support and international legitimacy for territorial conquest.

Mongkut’s discovery of the Stele was comparable to finding a lost Magna Carta at just the right moment. Siam’s rulers realized that written language was key to civilizational legitimacy in Western eyes, and the inscription provided exactly what they needed: tangible evidence of a tradition of literary sophistication that overturned European assumptions of primitiveness.

Diplomacy

In 1855, a new British special envoy to Southeast Asia arrived in Bangkok. Sir John Bowring was governor of recently conquered Hong Kong and represented a nation at the height of its powers. By this point, Britain controlled almost a fifth of the world’s land mass, produced over half its industrial output, and had the most powerful navy the world had ever seen. Siam, on the other hand, had been in self-imposed relative isolation from the West since 1688, and had recently been accused of dealing contemptuously with British and American trade missions. John Crawfurd, a British envoy, described the Siamese as ‘unusually sordid, insincere and rapacious’. Under Rama III, Siam’s king and Mongkut’s older half-brother, most people thought Siam was headed down the same path as its neighbors.

After Rama III died in 1851, Mongkut was crowned king, taking the name Rama IV. Mongkut was popular with the British, who thought him more friendly to their interests than the other contenders for the throne. As a monk, he had studied Western languages, sciences, and technology. He spoke English and Latin and had continuous contact with diplomats and Catholic and Protestant missionaries. He was fascinated by the Western concept of material and technological progress, an idea in stark contrast to Buddhist cosmology’s emphasis on repeated cycles of decay.

European diplomats had often found Asian rulers, including Mongkut’s predecessors, to be aloof and disdainful. Arriving in Siam for his meeting with the new king, Bowring would have expected more of the same. Instead, he found a ruler who displayed a sophisticated understanding of the Western world. King Mongkut received Bowring in his private apartments for cigars and wine. They bypassed interpreters and spoke in English. Bowring noted that the king’s library had a vast array of books ranging from Western science to novels by Sir Walter Scott.

British imperialists were unaccustomed to negotiating with Asian rulers who could meet them on their own intellectual ground. This was an era in which Thomas Macaulay, the colonial administrator and historian best known for his History of England, could write that ‘a single shelf of a good European library was worth the whole native literature of India and Arabia’. Bowring, however, was willing to believe the Siamese were cultured. His influential 1857 work, The Kingdom and People of Siam, featured the Ramkhamhaeng Stele, recognizing it as a major historical document that provided evidence of early Thai writing and was crucial to understanding Sukhothai’s sophisticated governance and society.

The treaty that Mongkut and Bowring negotiated was a decisive shift in the relationship between Britain and Siam. Three decades earlier, in 1826, a treaty had opened limited trade between the two countries. The 1855 treaty was by no means generous. It capped Siamese import and export duties at 3 percent and gave significant extraterritorial rights to British subjects, meaning Britons could not be tried in Siamese courts. But the treaty massively increased foreign trade, and crucially, Siam continued to be treated as a sovereign negotiating partner.

The contrast with other countries in the region is stark. Following its defeats in the Anglo-Burmese Wars of 1824–26 and 1852, Burma had been forced to surrender four provinces, pay a crippling £1 million indemnity (about $160 million today), and accept the presence of a British overseer with veto power over the king’s decisions. Similarly, China’s attempt to enforce a ban on Britain’s opium trade in 1839 had resulted in the loss of Hong Kong and the start of what nationalists would later dub the ‘century of humiliation’. Mongkut demonstrated what strategic compromise could achieve when resistance failed.

Xenophilia

Siam’s distinctive approach was not as simple as a cigar-smoking bibliophile king. That was a small part of a broader strategic performance of European norms, an attitude that set Siam apart from other countries in the region.

In response to Britain’s first embassy to China in 1793, Emperor Qianlong dismissed Western technology, declaring that the Middle Kingdom ‘possesses all things in prolific abundance’ and had no need for the goods of ‘barbarian merchants’. Chinese rulers were trained in classical Confucian thought that emphasized China’s cultural superiority, and their insularity and xenophobia made it easy for Europeans to justify their civilizing missions. Similarly, Vietnam’s Emperor Minh Mang rejected French overtures in the 1820s and 1830s. Vietnam’s isolation would later provide France with the pretext for intervention. Both rulers were products of educational systems that viewed foreign knowledge as contaminating rather than useful.

When Asian rulers did embrace modernization, they often did so reactively, after the West had already put them under substantial pressure. Despite having ancient scripts dating back to the sixth century, magnificent Buddhist temples, and sophisticated legal traditions, Burma was systematically dismantled by Britain between 1824 and 1885. The crucial differences between Burma and Siam were timing and agency. Burma was first invaded in 1824, before the ‘civilizing mission’ ideology was in full swing. King Mindon, who ruled Burma from 1853 to 1878, modernized extensively after Britain’s initial conquest, establishing telegraph lines, railways, and modern schools. Yet by then, Britain had generated military and political momentum that rendered Burma’s cultural credentials and reforms irrelevant. When Britain manufactured border disputes to complete Burma’s annexation in 1885, Mindon’s modernization efforts provided no protection against a colonial power that was, by that point, committed to conquest.

Japan also modernized reactively after the US Navy’s ‘black ships’, under Commodore Matthew Perry, forcibly ended the Tokugawa Shogunate’s policy of isolation in 1853. China attempted to do something similar following the humiliation of the Second Opium War, though with much less success. In comparison to its neighbors, Siam strategically adopted and demonstrated the markers of civilization that European powers used to justify colonization elsewhere from a position of relative strength.

In 1868, Mongkut extended invitations to British and French officials to observe a solar eclipse. This was a grand affair with pavilions and accommodation prepared for the attendees, the goal of which was to showcase Mongkut’s commitment to science. Mongkut accurately predicted the timing and location of the eclipse using a combination of traditional Thai timekeeping methods and Western astronomical techniques. According to some stories, his calculations were so precise that they trumped those of the French astronomers present.

The Siamese court knew what to perform to whom. France had long considered science to be the true marker of civilization, while Britain championed international law and free trade. Mongkut demonstrated Siamese civilization through measures that European powers used to justify colonization elsewhere.

Creating brand Siam

Mongkut did not live to see his diplomacy come to fruition: he contracted malaria at the eclipse event and died a few weeks later. His son Chulalongkorn (Rama V) was a minor at the time of his father’s death, but assumed full control of Siam in 1873, finding himself at the head of a multinational, decentralized, and dynastic kingdom.

Across Europe and its empires, traditional monarchies were adapting to an age of nationalism by imposing unified national identities from above. France had established a centralized administrative system with uniform laws, and aggressively promoted Parisian French as the national language, systematically suppressing regional languages like Breton, Occitan, and Alsatian. Meanwhile, Britain more gradually imposed standard English through education reforms and civil service requirements, displacing Welsh and Gaelic. The Austro-Hungarian Empire was, in its Magyarization policy of the 1880s, doing the same: imposing Hungarian language and culture on ethnic minorities like Slovaks, Romanians, and Croats to create a single national identity in the Hungarian part of the dual monarchy. Germany and Italy both transformed from complex patchworks of states to unified nations in the late nineteenth century.

Despite Europeans having only recently created their centralized nation states, this became the new standard for civilization. Many Asian powers lacked the cultural homogeneity, clear boundaries, and single language and government that nineteenth-century Europeans considered civilized. And these diverse and decentralized realms were ripe for divide-and-rule tactics. The British and French depicted non-dominant groups in realms like Burma, Cambodia, and Vietnam as oppressed minorities in need of their protection.

Nineteenth-century Siam was, in theory, vulnerable to the same tactics. It encompassed dozens of ethnic groups and lacked a unified Thai identity. Its people were governed by the mandala system, the traditional premodern governing system in Southeast Asia, its equivalent to Europe’s feudal system. Under the mandala system, rather than fixed borders, the king’s sovereignty faded with distance, with outer parts of the kingdom recognizing overlapping zones of authority from elsewhere.

When the British became neighbors to Siam’s south and west in the 1820s by acquiring the Straits Settlements and annexing lower Burma, their request to demarcate fixed mutual borders baffled the Siamese. European powers, once they understood the system, would attempt to exploit this difference in understanding. The French seized on Siam’s multiethnic structure, arguing that the Lao constituted a distinct race deserving protection from Siamese domination. They also claimed that the Siamese were not a true race, because they had become too intermixed with the Chinese.

Instead of giving in, Siam adapted through a program of nation-building. Chulalongkorn promoted the idea of ‘Thai-ness’, a single national identity that could accommodate Siam’s dozens of ethnic groups while presenting as a unified state to the Western powers. When confronted with French claims about distinct ethnic groups, Chulalongkorn countered: ‘The Thai, the Lao, and the Shan all consider themselves peoples of the same race. They all respect me as their supreme sovereign, the protector of their wellbeing.’

As well as a shared national identity, the people of Siam needed a shared language and clearly defined borders. The Bangkok dialect was elevated to ‘standard Thai’ and was spread through a newly created school system. The mandala system was dismantled in favor of a centralized bureaucracy, reorganized into administrative units that drew inspiration from the old mandala circles but imposed newly fixed territorial control. Modern mapping projects helped draw borders with unprecedented accuracy. A national census counted subjects who, a few decades later, were required to adopt surnames. Telegraph lines and railways created communication infrastructure that solidified Bangkok’s authority.

This approach differed dramatically from that taken by Vietnam. While Vietnam had a coherent identity dating back centuries, the ruling Nguyen dynasty struggled with adapting it to the Western concept of nationhood. Emperor Tu Duc was resolutely committed to maintaining Vietnam’s traditional Confucian order, and rejected reform. Vietnam retained a hierarchical system that privileged ethnic Vietnamese while marginalizing minorities such as the Cham and Khmer. The French capitalized on these self-imposed divisions by governing different regions separately – Cochinchina as a colony, and Annam and Tonkin as protectorates – thereby deepening regional fragmentation. The end result was an effective dismantling of the Vietnamese state.

By embracing the Western concept of nationhood, Chulalongkorn was able to force Europeans to recognize Siam as a sovereign country, rather than as a wilderness in need of conquest. Chulalongkorn embarked on grand tours of Europe in 1897 and 1907, amid sustained British and French pressure on Siam’s borders. Across Europe, he insisted on being received as a monarch of equal rank to his hosts. His entourage wore Western clothing, observed European court customs, and spoke fluent English. Some of his sons, including future Siamese kings, attended elite institutions such as Eton. The message was clear: Siam was a peer.

Modernizing Siam

As well as promoting the idea of a unitary Thai identity, Chulalongkorn took up a comprehensive program of modernization and centralization. In 1885, Chulalongkorn established the first modern public school. He created the Ministry of Education in 1892. By the end of the 1890s, the kingdom had about a hundred schools teaching a standardized curriculum, including the elite Suankularb Wittayalai school that trained future government officials in both Thai tradition and Western knowledge. It is still one of Thailand’s most prestigious secondary schools.

The legal system was also up for reform. The relatively unequal treaties imposed by Britain and France had granted extraterritoriality to European subjects, exempting them from Siamese law in a humiliating denial of sovereignty. Chulalongkorn’s solution was to create a legal system that Europeans would recognize as legitimate. He assembled a team of foreign advisors, most notably the Belgian jurist Gustave Rolin-Jaequemyns, who became a key architect not only of Siam’s new civil law code but of its entire diplomatic strategy. The choice of a Belgian was strategic: adjacent to British and French legal traditions but crucially not from either colonial power. Chulalongkorn built modern law courts and staffed them with the burgeoning administrative class trained in Siam’s schools. By demonstrating that Siam possessed a ‘civilized’ legal system, Chulalongkorn could argue for the abolition of extraterritorial privileges.

Chulalongkorn also Westernized the military. He hired German advisors and sent some royals to the Prussian military academy. In line with the Prussian model, military command was centralized under the monarch, noble levies were replaced with a standing army, and a professional officer corps was established.

Religion, too, was reformed to appear more rational by Western standards. Chulalongkorn supported the Thammayut monastic order’s reformation that made Siamese Buddhism appear less superstitious and more disciplined and text based. Even time itself was standardized. The kingdom adopted the Gregorian calendar and aligned its timekeeping with international norms. The quantification and standardization of time was foundational to Western modernity. Mechanical clocks, synchronized schedules, and precise measurement undergirded capitalism, science, and bureaucratic efficiency.

Cultural transformation reached into daily life through the promotion of siwilai, a Thai loan of the word ‘civilized’ that became central to the kingdom’s modernization program. Western-style clothing became increasingly common, often through official encouragement. Practices deemed uncivilized by European observers, such as chewing betel nut, were discouraged.

Perhaps most significantly for Western audiences, Chulalongkorn abolished slavery. Before the modernizations, Siam had a complex hierarchy of debt bondage, war captives, and corvée labor. By the late nineteenth century, slavery had been abolished in most of Europe and the Americas. Thailand followed suit. An 1874 decree freed children born into slavery when they reached the age of 21. It was followed by an 1884 ban on capturing and selling ethnic minorities into slavery. The 1905 Slavery Abolition Act went further, dismantling most remaining slavery and reforming debt bondage by crediting laborers’ work toward their principal debt rather than merely as interest. By 1908, slavery was criminalized and the government had a program of education to prevent former slaves from falling back into servitude.

In 1907, the French newspaper Le Figaro congratulated the king for his successful pursuit of modernization. It drew parallels between Siam and Japan, suggesting that France may now develop a beneficial relationship with an independent Siam. The same newspaper reminded readers: ‘one must not forget that the king of Siam, though his state is Asian, is a king of dignity like that of the Europeans.’

Soft power meets hard power

Siam’s strategy was tested during the Paknam incident in 1893. French gunboats forced their way up the Chao Phraya River toward Bangkok. Officially, the Paknam incident centered on a dispute between the Siamese and French regarding territory east of the Mekong River. In truth, it was prompted by France’s growing colonial ambitions following its conquest of Vietnam.

On 13th July, three French warships attempted to force passage past Siamese fortifications. The Siamese fired upon the warships. The French fired back, killing three sailors and breaking through the Siamese defenses. The French proceeded upriver towards the capital. They demanded that Siam recognize French Indochina’s claim to the disputed territory. In the face of overwhelming military force, Chulalongkorn ceded the territory, making a necessary concession to preserve the kingdom as a whole. Siam was forced to cede territory to the French in what is now Laos and Cambodia, creating much of the modern borders of Thailand. The French established a demilitarized zone along the Mekong, and the Siamese had to pay an indemnity of three million francs. European powers readily resorted to gunboat diplomacy but even here, France felt compelled to frame its demands in terms of legitimate Vietnamese territorial claims rather than naked conquest.

Between 1893 and 1909, Siam would cede around 40 percent of its territory to France and Britain. But the strategic retreat meant Chulalongkorn preserved what mattered most: the political independence of Siam’s core territories. The lost lands, predominantly Lao and Khmer-majority regions along the Mekong, had always existed at the outer edges of the mandala system, where Bangkok’s authority was weakest. Chulalongkorn specifically prioritized the Chao Phraya valley, the Thai-majority heartland, which was the source of the kingdom’s agricultural wealth and political identity. By surrendering territories where their control was already tenuous, the Siamese leadership prevented France from advancing into this core.

Siamese Wins

Some modern historians, many of them Thai, have questioned the authenticity of the Ramkhamhaeng Stele, suggesting that it may have been modified or fabricated to serve Siam’s existential myth-making needs. The Stele’s discovery mirrors other nationalist discoveries of the era, such as the Nibelungenlied, a medieval epic poem that German intellectuals promoted as a symbol of Germanic heroism and unity.

Whether Mongkut’s discovery was a true ancient artifact or a recent creation matters less than what Siam did with it. By the time French gunboats pointed their cannons at the Grand Palace in 1893, Siam had successfully established itself as a civilized nation worthy of sovereignty. The painful territorial concessions it made were closer to concessions European countries were in the habit of extracting from each other. It was not the kind of conquest that Europeans were doing in the New World, Africa, or the rest of Asia.

The kingdom that emerged from the colonial crucible bore little resemblance to the decentralized mandala system that had puzzled British border commissioners in the 1820s. It had become a Western-style nation. The ruling family of Siam had imposed one dialect across the populace, created a centralized state with the capacity to measure and monitor its people, standardized education, built infrastructure, and abolished slavery. They even created a national myth.

Yet rather than becoming Westernized by conquest, this transformation was carefully orchestrated from within. Mongkut and his son Chulalongkorn weren’t to know this, but their strategy may have eventually protected Thailand from a great deal of suffering. Weak institutions built by colonists fell apart after twentieth-century decolonizations, contributing to the deaths of millions of people across Cambodia, Vietnam, Malaysia, and Burma.

Siam demonstrated that a nation could transform itself on its own terms while others around it lost their chance.​​​​​​​​​​​​​​​​ The monk prince who discovered that stone in Sukhothai, armed with his synthesis of traditional Buddhist and Western scholarship, established a template for survival that would serve his kingdom for generations.

Derek Hopper is a writer and university lecturer based in Qatar.

Subscribe now

Martin Gould and Koji Flynn-Do explain why animals’ lives might, finally, be about to get better.

In September 2022, a new kind of pig farm opened in China: a twenty-six storey high-rise in Ezhou, near the Yangtze River. Months later, a second tower went up. Inside, pigs are kept by the tens of thousands on each floor, monitored by cameras and fed, watered, and moved with mechanical precision. More than a million pigs are born, raised, and slaughtered each year without ever setting foot outside.

These high-rise factory farms are only one example of a common pattern. As countries get wealthier, farming consolidates into fewer, larger, more brutal farms. Once average incomes reach around $1,000, chicken farming starts to industrialize. By the time a country reaches $10,000, the vast majority of chickens are raised in battery cages. In countries with average incomes above $30,000, almost all pigs are raised in warehouses.

Factory farming exists to meet rising demand for meat. Meat is tasty, nutritious, and much more expensive than rice or bread. As people get richer, they want more of it. When Park Chung Hee took power in 1961, South Korea’s meat supply was around four kilograms per person per year, less than the weight of two chickens. When his rule ended in 1979, that figure had quintupled. By 2000, it had increased tenfold. Today, the average South Korean eats more than 120 kilograms of meat and seafood combined per year.

Share

In poor countries, where inadequate nutrition is the norm, rising meat consumption is good for people’s physical development and health. Diet is only part of the story, but the pattern is unmistakable. In South Korea, average male height has increased from around 160 centimeters in 1961 (five feet three inches) to around 175 centimeters (five feet nine inches) today. Chinese men have seen a similar height gain: from around 165 centimeters before the country industrialized to around 175 centimeters today.

But industrially farmed animals suffer immensely for it. Chickens’ beaks are frequently sheared off and their legs are too weak to support their bloated bodies. Pigs have their tails docked and teeth clipped without anesthesia, and mother pigs live much of their lives in gestation crates too small to turn around in. These lives are not worth living. The question is whether such brutal treatment is inevitable.

For many of modernity’s worst problems, a combination of affluence, maturity, and political pressure has created self-correcting mechanisms over time. As countries got richer, things that initially worsened due to industrialization eventually improved. Will industrial meat production follow the same path?

The Kuznets curve

In his 1853 masterpiece Bleak House, Charles Dickens memorably described the air pollution of industrializing London: ‘flakes of soot … as big as full-grown snowflakes’. When Dickens was writing, London’s air pollution had been steadily worsening for decades as Britain industrialized, and it would continue to worsen for nearly half a century longer. But had he returned a century later, he would have found the dirtiness of London’s air to have halved, and by 2000, it would be less than a tenth as dirty.

Growing economies and populations have historically driven deforestation. But improvements in crop yields have halved agricultural land per person since the Green Revolution of the mid-twentieth century. The rate of global forest loss peaked in 1980, and forest loss typically reverses when average incomes enter the range between $5,000 and $15,000, as farming becomes more efficient and marginal farmland reverts to nature.

These trajectories are called Kuznets curves, after the economist Simon Kuznets, who suggested that economic inequality rose as countries first industrialized but then fell as they became even richer. His curves appear in many places. Wealthy areas clean up their local air, forests, and factories. Their inhabitants are less obese than those of industrializing countries, and have more leisure time and better working conditions.

But Kuznets curves don’t appear in all places, especially not when the industrial activity has no direct costs for the rich country. Emissions have fallen substantially in Germany and Britain, but much less in America and Australia, which remain some of the world’s biggest carbon emitters. The total amount of material people consume, and the amount of waste they generate, do not seem to decline but rather keep growing. The question is whether animal suffering is closer to child labor, and will decrease as we grow richer, or to waste generation, meaning animals will continue to suffer in ever-greater numbers.

Subscribe now

How much meat do we eat?

Start with consumption. In about ten European countries, including Germany, Austria, the Netherlands, and Denmark, meat consumption per capita appears to have peaked. Northern Europeans ate far more meat in the early 2000s than they do now.

Walk into a German supermarket today and the change is visible. The queue at the butcher counter has thinned; the same floor space now splits between marinated chicken, ready meals, and plant-based sausages. Pork, once the unquestioned centerpiece of German cuisine, has lost its grip: consumption fell over 10 percent just in the five years before 2025. The shift is even more noticeable in institutions. Forty years ago, a daily university menu would have centered on schnitzel and pork stew. Today, similar establishments compete to offer the best vegetarian and vegan options. In 2021, 34 cafeterias across Berlin’s universities moved to menus that were 96 percent plant-based.

But many rich countries aren’t like Germany or Denmark. Average incomes reaching $40,000 or $50,000 do not reliably herald a decline in meat eating. A more common story is a plateau. This is not bad news, as it implies that there is something of an upper limit on meat consumption. The plateauing countries, like Portugal, Spain, the United States, Britain, South Korea, and Australia, have all reached a ceiling of around 150 kilograms of meat per person per year.

Whether this ceiling holds may depend on shifting dietary fashions. Faddish new guidelines recommend eating two grams of protein per kilogram of body weight every day. If half of that protein were meat, the average American would need to double their meat consumption to meet the target. Drugs like Ozempic suppress appetite and reduce total caloric intake, but evidence suggests users shift toward protein-rich foods to preserve muscles, which could mean less food overall, but with perhaps a higher share of it meat and dairy.

More importantly, the kind of animal being eaten has changed. While Germans, Austrians, and Danes are consuming less meat, their diets are leading to the farming of more animals. In 1980, the average German diet required the slaughter of about ten farmed animals per year: roughly six chickens, one and a half pigs, and a third of a cow. By 2022, even though Germans were eating a fifth less meat, that number had doubled to twenty animals, largely driven by the shift to chicken and farmed fish: roughly 14 chickens, a bit less than one pig, one sixth of a cow, and six farmed fish.

From a welfare perspective, that makes a tremendous difference. Beef implies perhaps a day or two of suffering per kilogram, whereas chicken implies dozens of days of suffering per kilogram and farmed fish hundreds or thousands, depending on the species.

How we treat farm animals

Perhaps the continued high meat consumption in rich countries should be expected. The Kuznets curves for emissions of sulfur dioxide, particulate matter, and water pollution were not created by people consuming less. They were produced by people consuming differently. Cities increased the amount of heating in homes, but replaced coal with natural gas. People wash more laundry than ever, but the detergents no longer include phosphates. The way down the curve might not require people to eat less meat, but simply to produce it better.

For decades, the animal agriculture industry has focused on efficiency: cheaper feed conversion, faster growth, more animals per square foot. The result is that meat, eggs, and dairy are historically cheap. In 1946, the average American factory worker had to work more than half an hour to buy a pound of chicken. Today it takes four minutes. That single-minded focus on cost means there are now a lot of cheap, simple changes farmers can make to improve animal welfare.

Egg production is one example. For almost a century, farmers around the world have confined laying hens to battery cages offering less than a sheet of paper’s worth of floor space per hen. This comes at considerable cost to the welfare of hens, who cannot perform natural behaviors like spreading their wings, dustbathing, and nesting.

In wealthy countries, consumers and companies have shifted to producing and purchasing more cage-free eggs. Today, 45 percent of American, 62 percent of European, and 82 percent of British hens are kept in cage-free facilities. That amounts to hundreds of millions of hens spared from cages each year. Hens get two to three times as much floor space, which means each egg costs one or two cents more to produce. Richer consumers are willing to absorb those small costs.

In some countries and regions, cages for hens have been eradicated at the ballot box. Voters in rich places like Switzerland, Massachusetts, and California have elected to ban battery cages with resounding majorities. The Swedes didn’t even need legislation. By 2025, almost all hens in Sweden were cage-free thanks to retailers and shoppers rejecting caged eggs.

Other practices in husbandry are following a similar path. The European Union and United States now both require cattle and pigs to be stunned before slaughter, though exemptions exist for religious slaughter. Crates that confine pregnant sows and baby cows to spaces too small to turn around in have gradually been banned across the European Union, New Zealand, and several US states since the 2000s. Tail docking and castration without anesthesia, once standard for pigs, are restricted in the European Union and Britain.

New technologies can reduce suffering even further. In ovo sexing allows farmers to figure out the sex of chicks at the embryo stage, discarding eggs containing males. Adopting the technology means farmers avoid gassing or grinding up male chicks on their first day of life. The practice has already spared hundreds of millions of chicks from that fate in Europe, and has only just started spreading in the US.

Overall, these welfare improvements aren’t enough to offset the increased number of animals being farmed around the world. The sheer growth in the number of animals consumed, particularly chickens and fish, swamps the gains from better treatment. But in a few countries, the combination of declining meat eating and improved treatment is plausibly decreasing total farm animal suffering. New Zealand has banned battery cages, restricted sow crates, and mandated pre-slaughter stunning, and is also eating far less meat. The same may be true in Germany, Sweden, and a handful of other wealthy European countries. They offer a glimpse of what the other side of the curve might look like.

Subscribe now

Bending the curve

Income needs to reach a certain level before things can start to change. Whether you ask people in Sudan or Singapore, roughly the same share – between 70 and 90 percent – will tell you that they believe farm animals experience suffering and that their welfare matters. What changes with wealth is willingness to pay. In European surveys, 42 percent of the poorest quintile say they would pay more for higher-welfare products. Among the richest quintile, it’s 72 percent. No poor country has bent the curve: New Zealand has an average income of $45,000, Germany $55,000, and Denmark $65,000.

But income alone doesn’t guarantee progress. There are many countries that have surpassed these income levels and still have increasing meat consumption and less stringent farm animal welfare reforms than their peers. These include large economies like Japan, South Korea, and the United States.

Animal welfare improves fastest when the technology is already cheaply available. Stunning machines for chickens and pigs and cage-free housing systems for hens were cheap and effective before they started to be adopted at scale. Plant-based meat alternatives or lab-grown meat have yet to make any real dent in animal suffering: the entire alternative meat industry has yet to reach more than 2 percent market share for meat overall. This is easily explained by costs.

The only cultivated meat currently sold in supermarkets comes from a Singaporean company: a shredded chicken product that is 3 percent cultivated chicken cells and 97 percent plant-based, and retails at more than twenty dollars per pound. That is roughly ten times the price of conventional chicken and around five times the price of even premium products like pasture-raised chicken.

The failure of alternative proteins to make much of a difference in the past few years does not look so damning when we look at other Kuznets curves. For most of the twentieth century, leaded gasoline poisoned children’s blood across the developed world until lead levels plummeted within a decade of countries phasing it out in the 1970s. The air pollution that Dickens encountered steadily worsened for two centuries before its precipitous decline. Those curves began to bend when the problem became undeniable at the same time as a workable substitute existed. If indeed there is a Kuznets curve for animal suffering, it may take a tremendous amount of time and effort before we see the other side. What we see now is only a glimpse.

Koji Flynn-Do is chief of staff for US policy at LawAI.

Martin Gould is a Program Officer at Coefficient Giving.

Subscribe now

Medical innovation has transformed our lives, but hardly anyone realizes it. You can tell people that global life expectancy has doubled in a century and watch it fail to register – it just sounds like an abstract number, and numbers don’t convey how different life used to be. So let me describe it another way.

If you collapsed with a heart attack in 1950, hardly anyone around you would know how to help. CPR wouldn’t be invented for another decade. If you reached a hospital, a doctor would likely give you morphine, tell you to lie still and hope for the best: the prevailing view was that once the heart stopped pumping, nothing could be done.

That changed in the 1970s. Cardiologists began to understand that with each passing minute, more heart muscle would be starved of oxygen, and once dead, it wouldn’t be able to recover. But doctors could limit that damage by swiftly restoring oxygen supply with a clot-busting drug or a balloon catheter, opening blocked blood vessels and rescuing heart muscle.

In other words, the idea that a heart attack is something you could actively treat is only about fifty years old. Since then, the share of patients who have died within a month of having the most severe type of heart attack has fallen by two thirds.

Zoom out to cardiovascular diseases overall, and the trend is even more dramatic. People in rich countries today have roughly a quarter the chance of dying from cardiovascular diseases as they did in the 1950s at the same ages. Back then, there were no statins, no cholesterol testing, no bans on trans-fats, no anti-smoking campaigns, no bypass surgery, and no implanted pacemakers.

Diabetes, certain cancers, infectious diseases and other conditions have also become far more treatable than they used to be. A century ago, if you were diagnosed with type 1 diabetes, your remaining life expectancy would be a few months, and a ‘starvation diet’ would be one of your few options. Today, you’d simply need to wear a small device on your arm to track your blood sugar level and release insulin to stabilize it in real time. A century ago, it was unheard of for anyone to be cured of cancer, except, in rare cases, by a surgeon’s knife. Penicillin hadn’t been discovered, and a simple wound infection could result in sepsis and death. Smallpox still existed. It swept across the world and killed millions globally, year after year.

I think we underestimate how much medicine has progressed because we rarely experience life without it anymore. But it has added years, sometimes decades, to people’s lives. Maybe surprisingly, it’s also reduced inequality in life expectancy: we’ve seen greater gains at the bottom of the distribution, among those who would have died as children or young adults. At the same time, life expectancy has risen at every age.

The statistics go like this. In 1950, the shortest-lived 1 percent of babies born in France lived less than three months. But for babies born after that, that 1 percent started surviving longer and longer: up to six months for those born in 1960, three years in 1970, 18 years in 1980, and 25 years in 1990. Alternatively, you can think about the cohort you went to school with. If you were born in France in 1900, only two thirds of your cohort would still be alive with you when you reached the age of sixty. But if you were born in 1960, 90 percent of them would.1 So here’s another way to think about the rise in life expectancy. It means that far fewer parents lose their children, fewer children lose their siblings, and we grow old with more of our friends and family still alive.

A technological revolution

I used to have the impression that medical breakthroughs were uncommon; sometimes happening by pure chance, sometimes by sheer determination. I now see it as a continuous stream of medical innovation every year.

In the last few years, we’ve had new antiviral drugs against HIV that can prevent infections with nearly 100 percent efficacy, with just a single dose given every six months. There are new drugs that reduce cholesterol levels by 60 percent for people already on statins. There are new treatments that slow the progression of cancers, including certain lung cancers, brain cancer, pancreatic cancer and multiple myeloma, by more than half. And over the past five years, we’ve had the first vaccines against four more diseases: Covid-19 of course, but also the first malaria vaccine, the first chikungunya vaccine, and the first RSV vaccine.

For most of our history, finding effective new treatments really was a rare occurrence. After Edward Jenner discovered the first vaccine, it took nearly ninety years for scientists to work out how to develop the second. People relied on plants and natural remedies for centuries, with little idea of which ingredients worked or why. Some of these remedies were effective, but contamination was common, and without knowing their active ingredients, they couldn’t be produced reliably at scale.

After chemistry matured into a scientific discipline in the nineteenth century, it became possible to isolate and purify compounds, synthesize new ones, and screen hundreds or thousands at once to find those that had medical effects. Chance discoveries could now be made systematically, and chemists discovered medicinal compounds in soil (yielding antibiotics like streptomycin, vancomycin, and tetracycline), fungi (statins and immunosuppressants like cyclosporin), and coal tar and dyes (sulfa antibiotics and early anesthetics like benzocaine). It was a valuable strategy when little was understood about the biological causes of diseases or the mechanisms of drugs.

By the late twentieth century, new tools, theory, and knowledge made it possible to develop treatments based on better biological understanding in an approach that became known as ‘rational drug design’. Scientists learned to decipher the structure of the proteins that drove disease and design molecules to fit into them and block them. They began to grow and study living cells in the lab, engineer bacteria and yeast to bulk-produce proteins (including insulin, clotting factors, and antibodies), and automate the screening of hundreds of thousands of compounds in days. It all became faster and possible at a far greater scale. Back when the Human Genome Project was completed in 2003, it cost $50 million and took six months to sequence just one person’s genome. Now it costs a few hundred dollars and takes under four hours. The tools have also gotten sharper. Over the past two hundred years, the resolution of the best microscopes has increased over ten-thousandfold. Until the 1930s, no one had ever seen a virus. Now we can see them down to their atoms.

Advances in AI are taking us further. Tools like AlphaFold can help predict the shape of proteins in minutes rather than years. Others are being used to design entirely new proteins that don’t exist in nature, including enzymes that catalyze reactions no natural enzyme can and proteins that target cancer cells and destroy disease-causing proteins within them. Some tools can identify hidden microbial genes that encode antibiotics but aren’t expressed under standard laboratory conditions. Others have uncovered ancient versions of today’s CRISPR enzymes, much smaller than the ones we use now but still able to protect bacteria from viruses. They have turned up entirely new classes of gene-editing tools that can precisely cut and paste long stretches of DNA, allowing them to fix multiple genes at once.

Much of the early research behind these breakthroughs takes place in academic labs and institutes funded by governments and foundations, before they are developed into drugs, manufactured, tested in large clinical trials, and brought to market by pharmaceutical companies. The first antiviral drug against HIV, for example, came from cancer researchers at the National Cancer Institute, who systematically screened hundreds of drugs to test whether any of them inhibited the virus before eventually identifying AZT. Burroughs Wellcome (now part of GSK) then ran the clinical trials to get it licensed. Similarly, the GLP-1 hormone was identified by academic researchers in the 1980s, but it was scientists at Novo Nordisk who developed drugs that mimicked it and chemically modified them to last long enough in the body to be used medically for diabetes and weight loss.

Competitive pressure generally causes firms to underinvest in open, early-stage research: if they publish it, other firms can use their findings without contributing; if they don’t, others can’t test, challenge or learn from their work. So private firms tend to focus on later-stage research, predominantly for diseases with a large commercial market, as they are better equipped to do the capital-intensive work of running large clinical trials, navigating regulators, and scaling manufacturing than academic institutions and foundations. Together, this process of exploration, discovery, development, and scaling has transformed once-fatal diseases into manageable, and sometimes even curable, conditions.

Medical innovation is one of the most valuable things that happens in a modern economy. Economists estimate that gains in life expectancy in the United States from 1970 to 2000 were worth about $95 trillion to Americans, with roughly half of that coming from reductions in heart disease mortality.

Share

Bottlenecks and broken incentives

There are almost certainly effective treatments in the pipeline today that won’t reach patients for years, sometimes decades, and it’s not because of the science.

There are few better examples than the malaria vaccine. Malaria is a very complicated disease: it’s caused by a parasite that changes shape multiple times during its life cycle, making it very hard to know what to target with a vaccine. So the first malaria vaccine, introduced just a few years ago, was an amazing scientific breakthrough. But it was actually developed in the 1990s. The researchers who developed it struggled to find funding to test it at every stage of the process. There was little commercial incentive. It’s not profitable to develop vaccines for diseases that affect people in poorer countries, without being able to recoup the R&D investment – even if millions of children might benefit and there’s a huge economic return. It took foreign aid and philanthropy to fund it, and it took literally decades to reach children who needed it, even as half a million of them were dying from malaria every year.

Those children aren’t coming back. But we can prevent this from happening again. Economists have come up with an idea to change the incentives with an advance market commitment, where donors promise ahead of time to buy a vaccine at a certain price per dose. The payout occurs only if it’s developed, and proven safe and effective. The commitment gives companies the confidence to invest, and can bring vaccines into existence that would otherwise never get made. Most importantly, it makes sure they’re manufactured at scale and sold at an affordable price, so they reach those who need them.

That idea was used over a decade ago to develop new vaccines for pneumococcal disease, a deadly bacterial infection of the lungs and brain. Pneumococcal vaccines already existed, but they didn’t include the strains common in Africa and South Asia. In 2009, several countries and philanthropists came together to fund an advance market commitment for new vaccines and it succeeded: several companies developed them and they reached children much faster than usual. It’s estimated that those vaccines have saved the lives of over 700,000 young children since then.

For all that, malaria is one of the better-funded global health diseases. Roughly $4 billion is spent on R&D across all tropical diseases combined each year, or around six cents for every thousand dollars of income in wealthy countries. Two thirds of it goes toward tuberculosis, malaria, and HIV. For the rest, the research pipeline is incredibly narrow, if it exists at all. One example is trachoma, a painful bacterial infection of the eyes that’s spread by flies. Repeated infections scar the eyelids, which gradually turn inward until the eyelashes scratch the eyes. Eventually, the scarring clouds the cornea, causing the person to go blind. Although we’ve made a lot of progress against new cases of the disease with antibiotics, around 400,000 people worldwide are blind because of the disease, and the pipeline of new diagnostics, treatments and vaccines in development for it is nearly empty, with R&D spending in 2023 falling to zero.

Many other tropical diseases, like leptospirosis, scabies, Buruli ulcer, mycetoma, and yaws, receive almost no R&D funding, and have research pipelines so thin they can support only a few dozen researchers worldwide. For some, like mycetoma, even the basic biology is mysterious, with little understanding of which molecular pathways drive the disease or how to target it with drugs.

The problem extends far beyond diseases of global poverty. Rare diseases, conditions affecting fewer than 1 in 2,000 people, are another example. Thousands of rare diseases have been described, and they collectively affect up to 6 percent of the population. But each disease is rare enough that the financial market for new drugs is limited and patients are often abandoned entirely; 95 percent of these diseases don’t have any approved treatment at all. Running trials for these conditions takes thinking differently: using smaller but more efficient trials, more coordination between researchers, and different approval pathways. For instance, many new gene therapies edit or silence specific harmful genes. Once researchers have found a safe way to get these therapies inside the cells where they’re needed, regulatory systems could treat the delivery method as a single platform, allowing the particular genetic sequence to be switched out to target different mutations of the same disease, or even different genetic diseases, without needing new approvals each time.

Or consider childhood leukemia. It used to be very difficult to test treatments for the disease because it is rare, and individual hospitals struggled to find enough patients to run clinical trials. So, from the 1960s onward, researchers built networks to recruit patients from across the US, and later Canada and Europe into larger clinical trials. That collaboration made it possible to learn what worked faster, and it’s why leukemia is no longer the disease it used to be. When I used to hear the phrase childhood leukemia, I’d picture a child who suddenly fell ill and whose parents were told they only had a few years left to live. That has changed. Before the 1970s, only 15 percent of children with leukemia would be alive five years after diagnosis. Now that figure is 85 percent. Most children in richer countries today survive and are effectively cured of the cancer.

These problems fall especially hard on tropical and rare diseases, but poor economic incentives mean we’re missing out on medical innovation across the board. Some of the biggest breakthroughs in medicine have been drugs developed for one condition that turned out to be effective for others, like GLP-1 drugs, which were initially approved for diabetes before being used for weight loss. Or SGLT2 inhibitors, which were also originally developed for diabetes, and have since turned out to be effective for treating heart failure and chronic kidney disease as well. This kind of ‘drug repurposing’, where companies test their own drugs for new uses, can be very profitable, but the incentive is time-limited: once a drug goes off patent and cheaper generics enter the market, the reward for running large expensive trials disappears and firms stop seeking approval for new uses. We don’t find out whether the drugs would have worked, and patients who might have benefited can’t access them.

Often, we may have many plausible candidates but no cheap way to test them. One way to fix this is through platform trials, which test multiple treatments for a single disease using a shared control group, which lowers the cost of testing each one and makes it possible to compare a range of treatments directly with one another. The RECOVERY trial in the United Kingdom, for example, tested over a dozen drugs as potential Covid treatments in a single trial over two years, at a fraction of the typical cost. Within a few months of the beginning of the pandemic, the trial identified that dexamethasone, a generic drug, was effective against Covid, cutting deaths by around a third among the sickest patients on ventilators. Its use is estimated to have saved hundreds of thousands of lives since then. The trial also ran much faster than usual, because participants were recruited through the UK’s National Health Service with simple, streamlined ethical approvals and consent forms.

The deeper problem is that technology isn’t always the barrier: sometimes it’s about the funding, the incentives, and the institutions. While genome sequencing costs have plummeted and the resolution of microscopes has surged, productivity in drug development has actually slowed down: the cost of getting a drug to market has been doubling in real terms every nine years for decades. This phenomenon was first described by Jack Scannell, who coined it ‘Eroom’s law’, the reverse of Moore’s law, which says that transistor chips roughly double in density every two years, allowing smaller and cheaper computers. The slowdown may be part of a broader trend: across many fields, including semiconductors and agriculture, the number of researchers has risen while the output per researcher has fallen and productivity has slowed down.

One explanation is that the lowest-hanging fruit is gone: many medicinal compounds in nature have already been screened, for example, and newer ones take more resources to discover. But given how much our tools of discovery have improved, I’m not convinced. More persuasive, I think, is the ‘better than The Beatles’ problem: new drugs not only have to be safe and effective, they have to be superior to the current standard of care, meaning they must clear a bar that rises with each generation of successful treatments. This isn’t just because older drugs already have some efficacy, but also because many have gone off patent and become generics, meaning they can outcompete new drugs on price as well. And as existing treatments improve and diseases become milder, it becomes statistically more challenging to confirm the benefits of new drugs, requiring larger or longer clinical trials.

It’s also become much slower and more complicated to run clinical trials. Before a drug can be approved, it must pass through successive stages of trials, with more participants and more data collected at each stage, before the data is submitted to regulators and approved for medical use. That overall timeline has grown from an average of six years in the 1970s to eight to nine in the 2010s.

Currently, each trial is built from scratch, with its own infrastructure, contracts and administrative overhead. Before a single patient is enrolled, each hospital or research center running the trial must review the protocol, negotiate contracts, obtain ethics approval and train staff; this process takes several months for each trial. Recruiting enough participants is then another challenge: trials seek out patients who meet narrow eligibility criteria, are willing to be randomized to treatment or placebo (meaning they may not receive the treatment at all), and will return to a clinic for follow-up visits sometimes dozens of times over several years. Around half of trials fail to recruit as many participants as planned; only one in five finish on time, with median delays of over a year. On average, it costs $54 million in the US, or around $37,000 per patient, to run a single late-stage trial for a single infectious disease drug.

The bureaucracy that surrounds trials today is largely a legacy of global regulatory guidelines introduced in the mid-1990s, which the trials industry interpreted defensively to minimize the risk of regulatory rejection. Even when regulators have since rolled back specific requirements, companies have been slow to change. Around a third of trial costs, for example, come from the practice of ‘source data verification’, where a consultant manually cross-checks every data point collected at trial sites against paper records, despite the FDA recommending against it because it catches few errors while adding huge costs. Because less than 10 percent of drugs that enter clinical trials are approved and the costs of development are large, the industry has remained financially risk averse, doing far more administrative work than is required, and far less efficiently than the process could allow.

That doesn’t mean we should sidestep the process of rigorous testing. Clinical trials are the best way to test whether drugs are actually effective, how effective they are, and what side effects they carry. Without them, we can’t reliably identify which drugs actually work, and we can easily be misled by random fluctuations, natural improvements over time, and selection biases that make drugs appear effective even if they have no benefit or actively harm patients. But the whole process could be much more efficient. It’s possible to speed up medical innovation without cutting corners or compromising on rigor, or better, through win-win solutions that improve both.

Subscribe now

A cure for every disease

Far too many diseases are still untreatable. It’s painful to imagine gradually forgetting your children’s names, forgetting how to hold a conversation, how to recognize your own home, and eventually even how to swallow food, while your family watches someone they love disappear. But that’s the reality for over thirty million people with Alzheimer’s globally, and we have almost nothing to offer them. Part of the reason may be because the brain is so complex, or because it’s lined by the blood-brain barrier, a tightly sealed layer of cells which keeps out most pathogens and toxins, as well as most drugs. That makes it challenging to design a drug that can cross the barrier, reach the right cells, and do what it’s meant to do without any unexpected side effects.

Cancer is not a single condition but hundreds of distinct diseases, driven by different genetic mutations, that can evolve rapidly and become resistant to each drug. Treating them often requires combination treatments, which attack cancers from multiple angles to reduce the chances of resistance developing, but many cancer drugs carry side effects that are highly unpleasant or even debilitating. Often, the only available options are chemotherapy or radiation therapy, which attack healthy tissue alongside tumors, causing lasting damage to the body.

I’m hopeful that progress is possible. Many conditions that once seemed untreatable have turned out not to be. Hepatitis C is now curable. Some diseases are now entirely preventable with vaccines, including cervical cancer and many liver cancers. Multiple liver diseases can be halted by new drugs that silence key proteins that drive their progression. CAR-T cell therapies, which reengineer a patient’s own immune cells to kill their cancer cells, mean that several blood cancers can now be put into lasting remission, and sometimes cured. New treatments have given cystic fibrosis patients lives that look much like anyone else’s, when they would have otherwise died in their twenties or thirties.

The most dramatic example is pancreatic cancer, one of the deadliest cancers, which has had a five-year survival rate of just 13 percent. It’s long been thought to be untreatable, because it is driven by mutations in the KRAS protein, which is notoriously hard for drugs to target because the surface of the protein is smooth, without an obvious pocket for drugs to bind to and block. In the last decade, however, breakthroughs in understanding the protein’s structure and interactions have helped develop new drugs that recruit a second protein to grab onto KRAS and block it. A new large phase three trial found that one such drug, daraxonrasib, roughly doubled survival compared to standard chemotherapy. Its side effects are rough, but knowing how to target the protein at all means scientists can build upon it. That work will be slow and difficult, but decades from now, we’ll look back and see the trajectory has changed, as it did for leukemia.

The list of bottlenecks is long, and we need to move faster if we want to solve them before it’s too late for too many people. It takes years to find and recruit patients for clinical trials and clinical trial infrastructure is highly fragmented. Commercial incentives mean that firms have little desire to develop drugs for many diseases whose patients are too few or too poor for it to be profitable, and there’s little incentive to repurpose drugs for new uses once patents run out. People have already come up with some ideas to solve them, including platform trials, advance market commitments, prizes to incentivize drugs for neglected diseases, risk-based monitoring of trials to focus on the highest-risk elements rather than on cross-checking every data point manually, large simple trials that operate through existing electronic health records, and continued randomization after a drug is approved to learn which dosing strategies, treatment durations, and drug combinations work best. But we also need new ideas, and people to test them, scale them up, and see them through.

We can make progress even in the hardest of cases. Ebola vaccines are incredibly challenging to develop through the regular route for many of the same reasons. There was little commercial incentive: Ebola largely affects the poorest, and its outbreaks are unpredictable, which makes it hard to set up trials in time. But what’s amazing is that we have a vaccine for the Ebola Zaire virus. Just over a decade ago, Gavi created an advance market commitment to fund it, and scientists came up with a different way to run clinical trials: ring vaccination. They waited for individual cases of Ebola to appear, and then quickly vaccinated the contacts of affected people, testing how well the vaccine prevented new infections. The World Health Organization coordinated the trials, the US government helped fund the vaccine’s development through BARDA and the Department of Defense, and once the vaccine was licensed in 2019, Gavi, WHO, UNICEF, and the Red Cross set up a global stockpile so doses would be ready for another outbreak. That stream of innovation, I’ve realized, is the result of countless people who thought differently, whether they were scientists, economists, managers or operators. Some of them changed their lives to make it happen. What they recognized was something quite simple: that diseases are not a fact of life, but problems that can be solved.

I’ve been writing about global health and medical innovation for years, and when I tell friends and family how much progress there’s been, the reaction is almost always the same: they had no idea most of it was happening. But progress isn’t inevitable, and breakthroughs don’t just come about when we have the tools to make them; they depend on how innovation happens in the real world. Right now, those processes are bottlenecked by funding, institutions and incentives. But we can change them. I think it’s more important than ever that we do so, because of cuts to spending on science, global health and foreign aid; we need to find ways to make our resources go further.

The central drama of the modern world is innovation against suffering, and we have made incredible progress so far. But we are doing far worse than we could. Millions still suffer from untreatable diseases, because of bottlenecks that we could fix. Fixing them is one of the most important things we can do. You can help make it happen.

Saloni Dattani is an editor at Works in Progress, writes the newsletter Scientific Discovery, and is an advisor to Coefficient Giving on clinical trial reform.

Share

In June 1973, chestnut colt Secretariat ran the Belmont Stakes so fast that no horse has matched him since. In less than two and a half minutes, Secretariat flew along the track, finally pulling away to win by 31 lengths, nearly three quarters of a football field. The other horses were so far behind they weren’t even in the television frame. Fifty years later, billions of dollars have been poured into breeding and training, yet Secretariat’s record still stands.

This is the paradox of the thoroughbred: breeders have spent centuries trying to create a faster horse, but in doing so, they may have bred away the very genetic diversity that would allow the horse to keep improving, or even to remain robust.

The paradox is not confined to the racetrack. The horse began as a terrier-sized, multi-toed forest dweller and, over 55 million years of evolution and a few thousand years of human intervention, became everything from the compact Icelandic pony to the streamlined racehorse. For most of that history, natural selection shaped horses slowly. Only recently has human ambition taken over, breeding animals for speed, endurance, and docility. The result is extraordinary diversity in form and performance, but also a dramatic narrowing of genetic diversity.

The first horses

The story begins 55 million years ago, deep in the warm, humid forests that covered what is now North America. You’d be hard pressed to recognize the earliest equid. Known as Hyracotherium, it stood only 14 inches tall, closer to a terrier than a modern horse. With four toes on the front feet and three on the back, it was ideally suited to moving gingerly over soft, marshy ground and nibbling fruits and leaves.

When global temperatures rose, between 18 and 15 million years ago, the equid family diversified into more than twenty genera (the subdivision above species living in woodlands and savannas). Most of these lineages have long since died out: today, only the genus Equus survives, which includes horses, zebras, asses, and donkeys.

As grasslands spread, over millions of years, equids developed into the horses we see today. Few evolutionary stories are as well documented. In the nineteenth century, when paleontologist OC Marsh unearthed a rich collection of fossil horses in the American West, their significance was immediately recognized. Here, laid out almost step by step, was evolution in action: horses transitioning from small, many-toed forest-dwellers into large, single-toed animals that grazed on grass rather than eating fruit and leaves.

That single toe, of which the hoof is the toenail, is the evolutionary innovation key to horses’ success. Life on the plains favored animals able to run efficiently over long distances. Rigid, lightweight hooves help with this: they minimize sideways wobble while storing and releasing energy like a spring, much like the shoes a human sprinter might wear, or even the prosthetic running blades sometimes used by amputees.

Over millions of years, the number of toes fell gradually from several to just one. Between 43 and 33 million years ago, Mesohippus still sported three toes of roughly equal size, with each touching the ground. By the time of Hipparion, about 23 million years ago, the middle toe carried most of the animal’s weight: the outer two toes had shrunk into vestiges, and the middle toe had become dominant. Finally, Pliohippus took a pivotal step toward the modern form, with a single weight-bearing toe ending in a solid hoof. This evolutionary streamlining produced the feet of today’s horses: a single, sturdy keratin hoof perfectly suited for galloping over open ground.

The route to the modern horse involved several detours. Hyracotherium, the progenitor of the horse family, appears in the fossil record in North America from around 55 million years ago, disappearing again 8 million years later. Around 12 million years ago, Hipparion, which, though three-toed, otherwise resembled modern horses, colonized Eurasia. Hippidion, which had short limbs and distinctively shaped facial bones, roamed the plains and mountains of South America from around 3 million years ago. Then, around 2 million years ago, the most recent common ancestor of present-day asses and zebras crossed the Bering land bridge formerly joining Alaska and northeastern Siberia.

Within the next half a million years, their descendants rapidly expanded across Eurasia and entered Africa at least two separate times. The descendants of the first migration diversified into various zebras, while the second led to modern donkeys and African wild asses.

Although the Americas were home to the earliest proto-horses, horses (including Hippidion and others) later became extinct in the Western Hemisphere around 10,000 years ago, alongside other megafauna like giant sloths, human-sized beavers, and armadillos as big as cars. All horses in the Americas today, including the free-roaming mustangs of the American West, come from European horses introduced after the Spanish conquest – descendants of the branches of the horse family that crossed into Eurasia and avoided extinction.

Share

The first horse tamers

Exactly when, where, and how horses were domesticated remained a mystery until recently, because fossils of wild and early domestic horses look very similar. For decades, archaeologists pointed to the Botai people of northern Kazakhstan, who lived around 5,500 years ago, as the original domesticators. Excavations revealed corrals, possible evidence of harnesses, and even pottery with residues of horse milk, a clear sign that the Botai people were not merely hunting and eating wild horses but had integrated them into many aspects of their lives.

While the Botai horses seemed like the obvious ancestors of modern horses, DNA told a different story: when scientists sequenced the genomes of twenty horses from the Botai site, they were not genetically similar to modern domestic breeds. Instead, Botai horses grouped closely with the Przewalski’s horse (pronounced shuh-VAHL-ski), a species discovered roaming the Mongolian steppe in the late 1870s and long considered the only truly wild horse.

This means that, contrary to the long-held belief that Przewalski’s horses were the last surviving truly wild horses, they are instead the feral relatives of the earliest known domestic horses found at Botai. While the Botai likely did domesticate horses first, their culture died out and their horses returned to the wild.

From steppe to stable

The horses we know and love came from a different source. They emerged more than a thousand years later, about 4,200 years ago, in the Pontic-Caspian steppes of what is now southern Russia, the same region that gave rise to famous horse-riding warriors like the Cossacks, the Golden Horde, and the Huns.

Scientists have worked this out through genetic analysis of ancient horse remains, combined with carbon dating to establish their precise age. A foal inherits half of its DNA from each of its parents. This is passed on with almost incredible accuracy, with around 2.7 billion base pairs (the data units of DNA) being copied correctly, but even so a few dozen errors creep in each time. A scientist can take the DNA of an ancient ancestor from a fossil, compare it with the DNA of a modern descendant, and then estimate how many generations have passed by counting the number of mutations that have accumulated.

In 2021, scientists did just this, sequencing genomes from 273 sets of bones ranging from 50,000 to 200 years old. By looking at the mutation rate and carbon dating the bones to confirm their exact age, they could calculate how quickly generations passed. About 4,200 years ago, the generational clock began to tick twice as fast, halving relative to wild horses, suggesting deliberate breeding.

Although bones can’t tell us how ancient horses behaved, their genomes give us clues. The researchers found that certain genetic variants became unusually common in these steppe horses. Such a pattern signals strong selective pressure: something in the environment consistently favored animals carrying particular versions of genes.

One region under heavy selection lies near a gene called ZFPM1, thought to play a role in the development of brain cells involved in mood, fear, and aggression. This suggests that early breeders might have been either deliberately or indirectly choosing animals that were more docile, allowing training, handling, and eventually riding. A second region under strong selection sits just upstream of GSDMC, a gene linked in humans to spinal problems such as chronic back pain and narrowing of the spinal canal. The signal of selection at this locus in horses suggests that breeders may have favored genetic variants associated with stronger, more resilient backs. As horseback riding spread, horses that could bear weight without developing spinal pathologies would have been more desirable.

These steppe horses, known to geneticists as DOM2 (that is, domesticated lineage two), spread like wildfire. By 2200 to 2000 BC, the DOM2 genetic profile spread beyond its region of origin to Anatolia, the lower Danube region, and Bohemia.

Archaeological evidence suggests that as DOM2 expanded, the populations of other wild horses gradually dwindled. By around 1500–1000 BC, almost all had been either absorbed or replaced. The tarpan horse, which roamed Eastern Europe, became extinct in 1909. By 1969, Przewalski’s horse was declared extinct in the wild. All living Przewalski’s horses stem from just 15 individuals captured around the year 1900. Only in 1992 did the descendants, numbering about 2,000, begin to be reintroduced on the Eurasian steppes.

The many so-called wild horse populations found around the world, including mustangs, Sable Island ponies, and Australian brumbies, all descend from modern domestic horses that became feral again within the past few centuries.

Horses for courses

Over time, humans developed distinct horse breeds with a wide range of phenotypes, from the tiny Argentinian Falabella, barely reaching a meter tall at the shoulder, to the towering French Percheron, capable of hauling tons; from horses with the standard gaits (walk, trot, canter, and gallop) to horses able to perform alternative gaits, like ambling, with no bounce, making them more comfortable to ride. The Icelandic horse’s tölt is so smooth that a rider can carry a full glass of beer without spilling a drop.

This diversity is the product of millennia of human intervention, selection, and preference, which left distinct traces in the horse’s genome. Compared with many other species, horses have a lot of diversity in the mitochondrial genome, inherited from the mother, but extremely limited variation within the Y chromosome, passed down from father to son. This contrast suggests a pattern in domestication: many local mares (female horses) were bred with a fairly narrow pool of stallions (males), concentrating paternal lineages while preserving variety in the maternal line.

Ancient DNA has given us a window beyond fossils, allowing us to track how phenotypes changed in ways that don’t show up in the fossil record. After humans, horses are one of the species ancient genomics has studied the most. In fact, the very first successful recovery and analysis of ancient DNA was from 150-year-old tissue from the extinct quagga, a close relative of the zebra, in 1984 – a year before the same was achieved with ancient human DNA. One of the oldest genomes ever successfully recovered and decoded comes from a horse bone preserved in the Yukon permafrost for over half a million years.

Coat color offers another window into early horse breeding. In living horses, the genes behind colors and patterns, like leopard spotting, tobiano patches (like the coloring of a cow), and cream dilutions (that lighten the color of a horse’s coat), are well understood. Genetic variants for dun (gray-gold, or tan), black, and leopard spotting appear in wild horses from the Upper Paleolithic, long before domestication began.

In the early Bronze Age, leopard-spotted horses were abundant at Kirklareli-Kanlıgeçit in present-day Turkey. But leopard spotting comes at a cost: horses with two copies of the gene coding for it have night blindness (the inability to see in low light). Its prevalence in certain ancient populations suggests breeders made deliberate choices, where aesthetic or symbolic value outweighed the trait’s practical drawbacks, to favor certain traits unlikely to persist in the wild. Tastes shifted with time: leopard-spotted horses lost their appeal after the early Bronze Age but gained favor in some regions during the Iron Age. And while spotted coats were popular in antiquity, solid chestnut horses prevailed in medieval Western Europe.

In the Middle Ages, horses were inseparable from power and status. Chivalry itself shares an etymological root with the Old French chevalier, literally ‘horseman’. Breeding and maintaining horses became a major industry of nobility and royalty. But they didn’t think in terms of fixed breeds the way we do today. Medieval writers grouped horses by function: the destrier trained for combat, the courser valued for speed, and the palfrey prized for a comfortable ride. Archaeological evidence backs this up. A large study of horse remains from thirteenth- and fourteenth-century London shows a range of sizes and builds, not the uniformity that we would expect from tightly restricted breeding.

Contrary to the modern idea of the medieval warhorse as a massive, towering beast, measurements from excavated horse bones across England show that most stood around 13 to 14.2 hands high, roughly the size of a sturdy modern pony. Even horses associated with elite or military contexts rarely approached the heights of today’s heavy draught breeds. What medieval riders valued instead were animals that were compact (meaning cheap to feed), strong, and agile enough to handle long distances, uneven terrain, and the demands of warfare.

Subscribe now

The need for speed

By the seventeenth century, horses had been serving human purposes for thousands of years, plowing fields, carrying soldiers, and pulling carts. But in early-modern England, one goal came to dominate: speed. In the 1660s, King Charles II, a fervent horse racing fan, set up a horse breeding and racing center at Newmarket, near London. While horse racing had been popular in previous centuries, Charles II and other racing enthusiasts were the first to methodically breed horses at scale to improve their race performance.

Breeders discovered that horses from Arabia, Central Asia, and North Africa were often faster: they were lighter, had a greater drive to run, and were better able to dissipate heat, allowing them to sustain high speeds without overheating. From three exceptional horses imported to England, Byerley Turk (imported in the 1680s), Darley Arabian (1704), and Godolphin Arabian (1729), breeders created the ‘thoroughbred’, a new type of horse designed for speed. Today, 95 percent of all living thoroughbreds are thought to descend from just one of these stallions, Darley Arabian, through his phenomenally successful great-great-grandson, Eclipse.

Horse breeding used to be guesswork: breeders paired fast stallions with fast mares and hoped for the best. But in 1791, the General Stud Book started recording every thoroughbred’s ancestry, turning pedigree into data. Breeders could now track bloodlines, deliberately plan matings, and see which combinations actually worked. Thoroughbreds today are defined as horses whose ancestry can be traced entirely to horses listed in the General Stud Book.

Harder, better, faster, stronger

Racing and breeding evolved side by side. Races in the seventeenth and eighteenth centuries involved multiple heats over distances of two to four miles, favoring horses with stamina. Horses didn’t race until they were five or six years old. But by the late eighteenth century, new races like the Derby and the St Leger reset the terms of competition. Horses now raced at three years old, over shorter distances of one to two miles.

This new format meant an increased premium on speed and precocity, and it has persisted, particularly in the United States and Australia. In Britain and Ireland, the tradition of long-distance races held on in parallel, leading to distinct types: horses that specialize in explosive speed at a young age, classic middle-distance horses, and horses bred for the stamina needed to excel in longer races.

After the creation of the thoroughbred and the Stud Book, Anglo-American breeders hit what appears to be an upper limit for speed surprisingly quickly. From the middle of the nineteenth century, when timing and record keeping became good enough to make reliable comparisons, race times improved steadily for about half a century. But although there was consistent improvement, it was small: times only fell by 1 to 2 percent, or 2 to 4 seconds over a 1.5-mile race. Race times for prominent races have not significantly fallen since around 1910. Secretariat’s 1973 Belmont time eclipsed the previous record by a stunning 2.6 seconds. However, the best time since then (Easy Goer in 1989 at 2 minutes 26 seconds) was a full 2 seconds slower than Secretariat’s.

Outside the elite, there has been continued improvement: one study found that between 1997 and 2012, accounting for differences such as ground softness and the number of runners, the average British thoroughbred increased its speed by 0.011 yards per second per year. So while we have likely reached the ceiling for how fast an exceptional horse can run, the median horse is beginning to close in on that limit, too.

The genes for speed

Until 2007, the Stud Book was the cutting edge of horse breeding technology. That changed when scientists sequenced a horse genome for the first time, that of a thoroughbred mare named Twilight. This allowed researchers to study the genetic basis of variation between individual horses far more systematically.

Another step forward in understanding thoroughbred genetics came with a paper published in 2010 that investigated the gene in horses that codes for myostatin, a protein that regulates muscle development. Mice, cattle, dogs, and even humans with mutations in this gene grow outlandishly large muscles because the gene normally acts as a brake on muscle development; with that brake damaged, muscles continue to grow unchecked.

The gene coding for myostatin in horses has around 1,100 base pairs. Looking at the exact sequence of this gene, called MSTN, in 148 thoroughbreds revealed that horses could have three different versions of the gene, differing by just one of those base pairs (either a letter ‘C’ or a ‘T’), and that each version was associated with success in different race formats. Horses with two copies of the ‘C’ version tend to be the best sprinters, those with two ‘T’s tend to be high-stamina long-distance runners, and those with one of each tend to fall between the two running profiles.

Later research found that variations in MSTN correspond with different muscle physiology. Horses with the ‘C’ version have a higher proportion of fast-twitch muscle fibers, which enable the explosive bursts of speed required for short-distance sprint racing, while those with the ‘T’ version have a higher proportion of slow-twitch fibers, which allow sustained effort.

Analyzing the DNA of modern horses and museum specimens then showed that the foundation stallions in the Stud Book all had two copies of the ‘T’ variant, and that the ‘C’ mutation probably entered through local British mares. This new variant entered thoroughbred genomes during the late eighteenth and nineteenth centuries, as English races like the Derby and St Leger got shorter. But the genetic trend is less pronounced in Britain than the United States, where races are rarely longer than 2,000 meters, and especially than in Australia, which favors 1,000 to 1,400-meter lengths. Commercial tests for this gene are now widely used by breeders and trainers.

The price of success

Horses’ genetic diversity remained large for thousands of years, before collapsing after the introduction of modern breeding. In nature, diversity is maintained by competing selection pressures that balance one trait against another. A wild horse does not just need to be fast. Its bones need to last. It needs to be resistant to disease. Even a domesticated working horse would, before intensive breeding began, face natural constraints on its traits.

In modern thoroughbred breeding, celebrity stallions with the best race records are in high demand. Their owners charge hundreds of thousands of dollars in stud fees, and they can sire perhaps a hundred foals per year. This means that the genes of particular males can quickly sweep to dominance in a population in a way that would be very unusual in nature.

Along with this loss of diversity, modern horses have accumulated harmful mutations as a result of breeding within a small population. In horses as in humans, repeated inbreeding concentrates harmful traits as parents who are related to each other are more likely to share the same mutations, resulting in offspring with two copies of the mutated gene, rather than one mutated copy and one normal copy, which could mask the mutation’s effects.

A genetic analysis of all Australian thoroughbreds, more than 135,000 horses, showed that inbreeding correlates with slower, shorter, and less lucrative careers. Amazingly, ten heavily crossed eighteenth-century ancestors account for over 80 percent of inbreeding in Australian thoroughbreds. Breeders are not blind to these risks. The American Quarter Horse Association, for example, requires breeding stallions to be tested for several genetic diseases.

As genetics has played such a large role in breeding horses, it is small wonder that horse breeders are moving with the frontiers of science. The world’s first cloned horse, called Prometea, was born in 2003. Horse cloning is now big business, despite an American Jockey Club ban, as other equine sports, including show jumping and polo, allow clones.

In 2024, the Argentinian company Kheiron announced the production of the first CRISPR-edited horses: Polo Pureza, an award-winning mare, was the foundation for five gene-edited clones with increased explosive speed. According to reports, the company modified the MSTN gene, so that the animals would develop more muscle fibers and potentially run faster as a result. Even so, the prospects for exceeding current records through gene editing are slim. Today’s racehorses are already tightly optimized for speed, balancing strength against the cost of carrying additional muscle mass. Adding more muscle is relatively low-hanging fruit, and if this were the key to unlocking greater speed, selective breeding would likely already have brought it about.

The thoroughbred paradox

At the Belmont Stakes in 1973, Secretariat ran in a category of his own. His extraordinary performance seems to have been the result of many unlikely factors aligning. After he died, Secretariat’s heart was reportedly measured at almost 10 kilograms – twice as much as normal. This appears to have been a physiological fluke, however, as it isn’t a trait reliably passed down to offspring. Combined with his unusual stamina, exceptionally powerful legs that allowed him to cover 25 feet per stride, favorable race conditions, and a skilled jockey, this produced a race time that has still never been matched half a century later, despite advances in breeding, training, nutrition, and veterinary care.

Secretariat offers a glimpse of the upper limits of biology. But as racehorses have pushed against that limit to become faster and more powerful, their population has become more fragile as the gene pool has narrowed. Our 4,000-year pursuit of the ultimate horse may not have much further to run.

Sophie Fessl is a science journalist.

Western countries used to let landowners build almost anything anywhere. Now, strict rules freeze the character of most neighborhoods, banning change and densification. Why? One theory is that modern buildings tend to be ugly, leading residents to support rules that exclude them from their neighborhoods. If this is true, then designing beautiful buildings could lead to more development. Housing reformists have vigorously debated this theory. Many British reformists are sympathetic, whereas Australian reformists tend to be skeptical, and Americans come down on both sides.

I do believe that ugliness plays a role in hindering development, but not in quite the way usually proposed. Ugliness is probably not a major driver of opposition to development by local people. Local opponents have many other concerns that are at least as important, like disruption from construction, loss of green space, pressure on local services, and diminished social exclusivity. But there is strong circumstantial evidence that ugliness is sometimes a major driver of opposition among people who live nowhere near the development in question. The first group are often called ‘NIMBYs’, for ‘Not in my backyard’. Let us call the second group ‘NITBYs’, for ‘Not in that backyard’.

While a few NITBYs are fervent campaigners, most are only casually interested. Their motives are faint but basically altruistic: their question is whether or not, based on a cursory assessment, a development is an objective improvement to its neighborhood. In certain contexts, the ugliness of a development can summon into being a huge cloud of NITBYs, whose interest in the scheme may indeed be superficial, but who can nonetheless be collectively potent.

Subscribe now

This model is especially compelling in the case of heritage conservation. Until the 1960s, the redevelopment of historic buildings was almost always permitted; today, it is almost always banned by conservation rules. Supporters of conservationism usually portray it as a response to the destructiveness of postwar planners, while opponents see it as a pretext for suburban NIMBYism. In fact, I shall argue, neither explanation is plausible. On the one hand, postwar planners were no more destructive than the Victorians or Georgians had been: the rate of demolition was historically normal, and cannot explain the wildly unprecedented policy response. And on the other hand, there was no need for NIMBYs to invent pretexts to ban new buildings, as they had already successfully prohibited the kinds of development they disliked in the first half of the century.

What had changed, and what generated the host of NITBYs that bore conservationism to victory, was the ascendancy of modernist architecture, a style that most people demonstrably find relatively ugly. Conservationism was a reaction to postwar redevelopment, as its proponents claimed. But it was less a reaction to what the modernists destroyed, than to what they created.

Ugliness and NIMBYism

Most debates about the policy significance of ugliness presuppose a certain model of how it affects housebuilding. On this model, local people resent the prospect of having ugliness inflicted upon them, they therefore oppose ugly developments, and politicians implement antidevelopment policies to placate them. This probably does happen sometimes: ugliness obviously does affect local people, and it would be surprising if this had no effect at all on their behavior. But a closer look suggests that ugliness plays only a minor role in motivating NIMBYism.

One reason for skepticism is that the occasional beautiful developments that do get built still often encounter fevered opposition. I’On (pronounced ‘ion’) in Charleston, South Carolina is one of the loveliest neighborhoods in North America. It has grown gradually since its founding in 1995, occupying a largely empty site between several existing residential subdivisions. Despite its walkable urbanism, tree-lined streets, and exquisitely detailed houses, thousands of local people petitioned against it when first proposed. The town council initially rejected the scheme, and a tortuous legal battle was fought to downzone the site through a local referendum. The NIMBYs were defeated only after the South Carolina Supreme Court found that zoning decisions cannot be made by referendum.

Such stories continue to appear today. The present king of the United Kingdom is an architecture enthusiast, and while he was prince, he orchestrated the delivery of several unusually attractive developments in his official landholdings (which, confusingly, are called the Duchy of Cornwall), including the famous developments of Poundbury and Nansledan. The Duchy is currently proposing a third, at Faversham in Kent. Its proposals show what would be one of the most appealing urban extensions in modern Europe. Nonetheless, it has triggered hundreds of letters of protest, as well as opposition from Faversham Council, the neighboring councils, and the local Member of Parliament.

Subscribe to print

Anecdotes like these are only weak evidence, not least because it is possible that the attacks would have been even more intense if the developments had been ugly. But they do suggest that beauty is not a panacea for NIMBYism.

A more powerful reason to doubt the role of ugliness in generating NIMBYism is that it gets the timing wrong. It is true that land-use rules became far more restrictive in the twentieth century. It is also true that the dominant architectural styles became much more unpopular: visual preference surveys consistently find that most people prefer traditional styles to modernist ones. Viewed from a distance, it looks plausible that the two are causally related: new buildings became visually displeasing to most people, and as a result they started to push back against them.

In fact, however, these changes happened in the wrong order. Restrictive zoning emerged in Germany and Austria-Hungary in the 1890s and spread quickly thereafter. Italian cities began introducing zoning plans in the 1900s, Britain created a national framework for zoning in 1909, and France followed suit in 1919. Zoning spread quickly across the United States in the interwar period. The restrictiveness of early zoning varied by country, but it could be very tight: in Germany and Austria most suburbs were capped at their existing densities well before 1914, and retain those densities today.

Subscribe now

Modernist architecture began at least a generation later. It emerged in the 1920s, chiefly in Germany, and spread in the 1930s as a minority style for clients who wanted something distinctively futuristic. In the United States, its impact was particularly faint, and famous American buildings of the 1930s and 40s, such as the Supreme Court Building, the Empire State Building, and Jefferson Memorial, remained resolutely Neoclassical or Art Deco. The global triumph of modernism came only in the 1950s.

This means that the Great Downzoning happened in countries where most new buildings were still in the traditional styles dominant in the nineteenth century, with at most a minority in the novel but decorative styles we now call Art Deco. No switch towards styles perceived as ugly had yet occurred: the general picture was one of continuity with the past. The flats that were banned by early zoning would generally have had classical cornices, pedimented windows, and rusticated basements. In other words, NIMBYism triumphed long before it became true that, in most people’s eyes, the average new building was uglier than the average old one. This suggests that, though making development beautiful might palliate NIMBYism somewhat, the effect would be modest.

Perhaps this should not surprise us. After all, even the most beautiful neighborhood only emerges after several years of extreme ugliness, while the buildings are under construction. And however beautiful buildings may be, development will still have its other well-known impacts on local residents, like loss of light and green space, pressure on infrastructure and services, increase in noise, damage to social cachet, and change in the character of the neighborhood. Removing the problem of ugliness may reduce NIMBY zeal a little, but the number of NIMBYs it satisfies entirely will be small.

Was conservationism a response to modernist destructiveness?

Ugliness is clearly not the main driver of local opposition to new developments. This does not, however, mean that ugliness is irrelevant to housing shortages in the West. Today, American prohousing advocates work to generate statewide or national support for development in order to override the locals, whose representatives control the strict zoning that slows down development and makes housing scarce and expensive. Prohousing advocates have secured some important victories. But one area where they have almost invariably failed is in enabling more development in historic areas. The same is true for reformist campaigns around the world. Old buildings are protected from redevelopment across the West, even when they sit on the most valuable central land.

How did conservation become so powerful? The two most popular theories are (a) that conservation rules arose as a response to the unprecedented destructiveness of postwar planners and developers or (b) that they are a clever pretext exploited by local NIMBYs. I believe the evidence contradicts both. Instead, heritage conservation arose once most people began to feel that new buildings were uglier than old ones, generating widespread opposition to the redevelopment of old neighborhoods by people who did not even live in them.

Subscribe now

The idea of conserving built heritage was not entirely a twentieth-century invention: the artist William Morris was an early proponent of the idea, establishing the Society for the Protection of Ancient Buildings, whose aim was to prevent the overzealous ‘restoration’ of churches and other historic buildings, in 18771. Morris was considered something of an eccentric in his own time, however, and architectural conservationism did not achieve anything like the popular support it has today until much later. In 1960, most European countries protected only a tiny minority of monuments from development, and even these limited restrictions dated largely from the 1940s and 50s. In most American states, the authorities protected nothing at all. Over the next twenty years, this changed drastically. Heritage registers were created, expanded, and strengthened. ‘Historic districts’ or ‘conservation areas’ were declared. Governments halted redevelopment schemes and redirected resources towards refurbishment and restoration.

Taken together, these measures were extraordinarily successful. In Paris, plans for the city introduced in the 1950s foresaw the demolition of two thirds of Paris’s city blocks over the next few decades, the bulk of Paris’s urban fabric from the nineteenth century. In Britain, the government actually did demolish at least 1.5 million homes in slum clearance schemes in the postwar decades. This was nearly 10 percent of the overall housing stock, but since these ‘slums’ were largely eighteenth or nineteenth-century buildings, it will have equated to between fifteen and twenty percent of the prewar stock. The United States seems to have had even higher rates of (privately led) destruction, with one report finding that nearly half of the 12,000 buildings recorded in the Historic American Buildings Survey of 1933 had been demolished or seriously mutilated by the middle of the 1960s.

Subscribe to print

Conservation policy brought such demolitions almost to a standstill. In France, 1.8 percent of pre-1949 dwellings were demolished between 1999 and 2013, an annual rate of 0.13 percent. German states report even lower rates of demolition of pre-1919 buildings. In Berlin in 2022, just nine buildings from before 1919 were demolished out of a total of 50,337, a rate of 0.018 percent. In North Rhine-Westphalia in 2023, 56 pre-1919 buildings were demolished out of 445,226, a rate of 0.013 percent. At this rate, it will take North Rhine-Westphalia 7,951 years to demolish its prewar stock entirely. In Britain, government statistics show a paradoxical increase in the number of pre-1919 dwellings between 2012 and 2024, the result of more homes being subdivided than demolished. Old American buildings are demolished rather more often, partly because they are timber-framed and therefore harder to maintain in safe condition. But in the American Northeast, the annual demolition rate of pre-1920 buildings between 1997 and 2001 was still only 0.25 percent, while in the West it was slightly over 0.3 percent2. In all these countries, the old buildings that are demolished are usually decayed and neglected. Many are in deprived post-industrial towns rather than in areas with strong pressure for densification.

The standard account of this shift sees the rise of the conservation movement as a response to the unprecedented destruction of heritage by postwar developers and planners. If we take the above figures in isolation, this seems plausible. The problem is that the destructiveness of 1945–1970 was really nothing new: earlier generations had destroyed just as much of their architectural inheritance without generating corresponding resistance.

A graphic illustration of this is visible on the Rue François-Miron in Paris, where stand Paris’s last two fully intact timber-framed houses. Until the sixteenth century, almost all houses in Paris were timber-framed. In the centuries since, Paris was never annihilated by a great fire, earthquake, or bombing, unlike Chicago, London, San Francisco, and Tokyo. But its tens of thousands of timber-framed houses were demolished for redevelopment, until, by the mid-twentieth century, only these two remained. Many of their sites have since been occupied by several progressively larger buildings, demolished and replaced in sequence.

This rate of redevelopment was not unusual. London’s medieval stock was mostly destroyed in the Great Fire in 1666. But the rebuilt City of London of the late seventeenth century has also vanished. Only somewhere between 100 and 200 homes survive from this period, though more than half a million people lived in the brick-built London of 1700. Some seventeenth-century buildings were destroyed in the Second World War, but the vast majority had already been demolished and replaced in the eighteenth, nineteenth and early twentieth centuries. Assuming 100,000 homes in 1700 and two hundred survivors today, that means a destruction rate of 99.8 percent, considerably more thorough than that of the Great Fire.

This process was perhaps even more marked in the United States, where the lack of height restrictions created intense incentives for redevelopment after the invention of elevators and steel frames in the 1880s. Fewer than twenty buildings of any kind survive in New York from before 1800. Boston has perhaps fifty or a hundred. Almost all of these buildings were already gone before the rise of modernism, demolished in the nineteenth century or in the furious building activity of the Roaring Twenties.

The rate of redevelopment could be even faster for certain types of building. In the Gilded Age, Manhattan had several hundred urban palaces, huge houses of the great business dynasties. Perhaps three quarters were demolished in a remarkably short period in the late 1920s. The Cornelius Vanderbilt II House was demolished in 1926 for the Bergdorf Goodman building, a department store. The Mrs William B. Astor House was also demolished in 1926 for the Temple Emanu-El of New York, a magnificent synagogue seating a congregation of 2,500 people. The Elbridge T Gerry Mansion was demolished in 1929 and replaced by the Pierre, a skyscraper hotel in a French Renaissance style. Ironically, all three of these buildings now have heritage protection, either as individual landmarks or as elements of historic districts.

This destructiveness was not limited to the private sector, especially in Europe, where medieval and early modern city centers frequently stood in the way of government schemes. The most famous case of this was ‘road cutting’, the policy of driving new arterial roads through city centers to improve traffic circulation, which was unnecessary before the age of the car on generously gridded New World streets. Many of Europe’s most famous urban streets were originally cuttings, like Regent Street in London and the Rue de Rivoli in Paris. Although cuttings mainly served a practical function, the tree-lined boulevards they created were also seen as aesthetic improvements.

There is one interesting exception to this destructiveness: urban railways. Nineteenth-century railways generated huge clouds of steam and coal smoke, and were regarded as blights on the neighborhoods they passed through. As a consequence, there are almost no cases in which a railway was cut through a city center: this is why cities like London and Paris have a ring of railway termini around the historic core. In other words, nineteenth-century people behaved just like twentieth-century people in respect of the one kind of infrastructure they found overwhelmingly ugly, just as the anti-ugliness theory would predict.

Subscribe now

Not everyone was happy about the tearing up of the old to make way for modern capital cities. People displaced by public works were obviously upset. Some commentators also regretted the destruction of historic buildings, especially when high-status buildings like palaces were replaced by low-status buildings like apartments. On the other hand, many Georgians and Victorians thought that redevelopment generally made their cities better. Although some ancient buildings were prized, normal houses generally declined in value over time, like cars or computers today: the default assumption was that buildings degraded with use, and hence that a new replacement was probably better.

Many people also thought that architects were getting better at designing attractive buildings. In 1868, The Builder, Britain’s leading architectural magazine, proclaimed that:

The overall picture here is one of striking continuity between the modernists and their predecessors. The modernists certainly demolished a lot, but not generally more than the Victorians or Georgians. In a sense, they were much less destructive, since they were halted when they had demolished only a substantial minority of the inherited building stock, while the Victorians and Georgians merrily continued until they had torn down almost all of it. The idea that postwar planners were moved by destructive animus towards the past is mostly a myth. Their views about urban change were historically normal, and had generally been shared by the wider public. The fact that redevelopment lost the consent of the public in the 1960s and 70s requires some other explanation.

Was conservationism a pretext for NIMBYism?

If postwar destruction did not cause the conservation movement, what did? Some readers will sense an obvious answer here. Housing reformers have long suspected that the conservation movement is mostly a pretext for NIMBYism. According to this theory, NIMBYs actually oppose development in order to exclude poor people or non-white people from their neighborhoods, or to sustain high property values, or from pure reflexive dislike of change. But since none of these is a persuasive argument for state coercion, NIMBYs therefore have to invent ‘pretextual’ reasons for opposing development. According to this theory, the idea that we must conserve architectural heritage is one such pretext.

The idea that conservation is a front for some other objection can seem plausible. When I attend planning committee meetings, I often have the overwhelming impression that opponents’ arguments are heavily pretextual: people who have never shown the slightest interest in heritage, and who have perhaps gutted the historic features of their own homes, suddenly become fanatical champions of some trivial historic object when it stands in the way of a development. (Intense interest may also be discovered in bats, newts, hedgehogs, bees, blackbirds, public housing tenants, flood risk, key workers, young people, old people, and so on).

Subscribe to print

As a general explanation for the rise of conservation in the twentieth century, however, the pretextual theory is surprisingly weak. The most obvious problem is once again sequencing: the rise of NIMBYism happened long before the rise of the conservation movement. As we have seen, suburban zoning spread across much of Europe before the First World War and then across North America in the interwar period, while conservation restrictions were extremely limited before the 1940s and only really became an important constraint on development in the 1960s and 70s. Many countries thus see gaps of forty or even seventy years between the two movements. Why didn’t NIMBYs invent conservationism earlier? Why, indeed, did they bother to invent it in the 1960s, long after they had already succeeded in protecting their affluent suburbs from development using other arguments?

The pretextual theory also struggles to explain the priorities of conservationism. As their name suggests, NIMBYs care chiefly about their own backyards – that is, typically, middle-class suburbs. Conservationism certainly does sometimes protect such places, but it is not restricted to them. It also protects working-class housing, commercial downtowns, public buildings, isolated country houses, and rural churches. In fact, it has often had far more effect on these buildings than it has on suburban areas, partly because suburbs were already protected by zoning, and partly because postwar governments wanted to clear ‘slums’, not middle-class neighborhoods. In other words, the conservation movement has the priorities of a NITBY movement, not a NIMBY one.

The difference between conservationist and NIMBY priorities is starkly illustrated by the divergent trajectories of city center districts. City centers with less historic build stock have frequently seen massive densification since 1950, because they have neither beautiful buildings for NITBYs to protect nor many local residents to object on NIMBY grounds. Thus many American, Canadian and Australian cities have seen forests of skyscrapers rising in their downtowns, while their vast suburbs, full of homeowning NIMBYs, remain petrified at low densities. Urbanists write about an ‘urban transect’, with densities gradually rising towards the center. These cities have developed a distinctive ‘L transect’, sheer at the core but flat everywhere else. This is the urban form that arises where local NIMBYism is a powerful force while NITBY conservation has less relevance.

By contrast, European city centers have been more perfectly frozen than perhaps any other kind of neighborhood: the building stock of central Paris, Rome, or Madrid has hardly changed at all since 1960. When European municipalities want to develop a skyscraper district, they generally have to find a low-end suburban site for it, like Paris’s La Défense or Madrid’s AZCA. If current plans for tall buildings around the Périphérique ring road go ahead, Paris may end up with a gentle U transect, with its highest densities displaced to the periphery. The obvious reason for this is that European city centers are where historic buildings are concentrated. They have no special significance for most NIMBYs – although they have often now gentrified, many centers were poor or depopulated in the 1960s – but they are the natural focus of NITBYs’ efforts.

Subscribe now

Where European cities have densified their centers since 1945, it is generally downstream of heavy bomb damage, as in Frankfurt and London. One famous study found that the Blitz ultimately made London’s economy 10 percent larger by enabling higher densities on former bomb sites. If London’s development restrictions were solely driven by NIMBYism, this would be inexplicable: sites where density had been randomly lowered to zero in 1941–1945 would if anything be expected to have lower density today, since the NIMBYs would have been resisting densification from a lower baseline. The reason that bomb damage enabled higher subsequent densities is that development is often blocked by NITBY protection of heritage architecture, not NIMBY opposition to density.

The fight against uglification

The true explanation of the spread of conservation is before our eyes: the rise of perceived ugliness. Redevelopment in the postwar era did not become more common or more destructive of heritage. But it did become, in the view of most people, uglier. This may have made people (even) more hostile to redevelopment near them. But its politically decisive effect was to make people more hostile to many kinds of redevelopment that did not affect them personally.

Most people are not principled opponents of redevelopment as such. Even today, developments on former industrial or logistics sites are often widely popular, like NOMA in Washington DC or Coal Drops Yard in London. But many people are principled opponents of redevelopment when it seems to uglify places. Very many people dislike the idea of their city or their country becoming uglier, even when they do not live near the development in question. They lend low-level, passive support to conservationist campaigns – support which aggregates into a powerful force in a democratic society.

How did people come to expect that new buildings would be uglier than what came before? As we have seen, this view was not standard historically. There are two main causes. First, in the middle of the twentieth century, traditional architectural styles were superseded by modernism, characterized by lack of ornament, repetitive facade patterns, and the use of exposed steel and concrete. This style emerged in the 1920s and became dominant for everything except suburban housing in the 1950s. As noted above, this style is broadly unpopular with the public. Replacement of prewar buildings in traditional styles with postwar buildings in modernist ones will thus, to most people, constitute uglification. This is the main reason that redevelopment switched from being aesthetically neutral to being uglifying.

Subscribe to print

Second, at around the same time, mass car ownership became dominant. Unlike modernist architecture, cars are basically popular. But it is nonetheless obvious that, although individuals benefit from cars, spaces are degraded by heavy vehicle traffic. This is irrelevant to how people feel about some kinds of redevelopment, but it is worth mentioning because of how it affects views of street cuttings. A nineteenth-century boulevard was an airy, tree-lined space along which horse-drawn carriages flowed, roughly at walking speed. On the whole, people seem to have preferred them to narrow side streets, as contemporary property maps illustrate. By contrast, nobody prefers living near a modern highway, to the extent that the authorities often build sight and sound barriers around them.

The anti-ugliness theory of conservationism lines up strikingly with the timing of events. Developers had been demolishing vast quantities of historic stock for thousands of years with barely a whisper of conservationist resistance; then, within ten or fifteen years of the ascendancy of an architectural style that the public disliked, an immense movement to halt the replacement of old buildings appeared almost from nowhere.

This theory also explains the priorities of the conservation movement. On the one hand, as noted above, conservationists have successfully protected a much wider range of pre-1914 buildings than NIMBYs have, including in city centers and other places where NIMBYs are scarce or absent. On the other hand, conservationists have been far less effective in defending postwar modernist heritage, being either indifferent to it themselves or unable to mobilize wider public support. There do exist conservation groups defending modernist architecture, like the Twentieth Century Society in Britain, and some high-profile modernist buildings have received conservation protection. But the numbers are tiny relative to prewar architecture. In England, 300 postwar buildings had been listed as of 2003, apparently growing to about 500 in 2015, out of some 375,000 listed buildings in total. The American National Register has a blanket ‘fifty year rule’, almost never including anything built within the last half century.

It also seems to me that the anti-ugliness theory is a plausible account of the emotional character of the conservation movement. It is true that many conservationists are not explicitly anti-modernist, and that today the orthodox high-status view is to support conservationism for old architecture and modernism for new ones. I suspect, however, that conservationists were appalled by uglification, but sought to avoid direct confrontation with high-status modernism by sublimating their appreciation of beauty into appreciation of heritage. Conservationists could thus promote traditional architecture without attracting charges of pastiche-mongering. In other words, many conservationists secretly love old buildings because they are beautiful, while pretending to love them because they are old.

Subscribe to print

Ugliness and the housing shortage

How much impact does conservation have on the housing shortage? The answer is complicated. Even in the European countries with the oldest build stock, only a fifth of housing dates from before 1919, while in the United States just 6 percent does. This suggests that rules protecting old buildings would do little to prevent the construction of new ones. But because historic buildings are concentrated in city centers, and because conservation systems tend to restrict what can be done in the environs of old buildings as well as within their curtilages, their effect is much greater than the raw statistics suggest. In fact, in most European cities, centers are effectively solid phalanxes of conservation designations, with new development largely displaced to inferior sites on the urban periphery.

The impact of conservation in the United States is smaller, but still considerable. Manhattan, central Boston, and San Francisco are not under blanket conservation designations like Paris, Amsterdam, Madrid, or Rome. But large areas still fall under conservation protections, often exactly the areas where YIMBYs would most like to enable development. This means that conservation, and hence ugliness, still acts as a significant constraint on development, albeit a less important one than the general rise of suburban zoning.

Cogent arguments may be given for the redevelopment of historic neighborhoods. Densifying central areas of cities creates new housing where it is most needed, and where it contributes most to economic growth. It enables people to walk and cycle to work, or to rely on public transport. New buildings could be safer, better insulated, and more environmentally sustainable than their predecessors.

For most Westerners today, however, the prospect of destroying their cities’ historic neighborhoods remains a horrible one. People feel implicitly that we possess a finite stock of beautiful buildings, and that they are more or less irreplaceable. When one is destroyed, we lose forever a part of our inheritance, and condemn our children to live in a permanently uglier world.

So long as people feel this, they will never tolerate the large-scale destruction of historic building stock, no matter what ingenious arguments YIMBYs make for it. Their dread of ugliness will hang over our cities like a curse, freezing them in the last moment of their beauty. If that curse can be broken, it will not be by the arguments of economists and transport planners, but by confidence that change is not disfigurement, and that it will yield cities we can love.

Samuel Hughes is an editor at Works in Progress.

Artwork by Daniel Cox.

Some of America’s most famous land was reclaimed from the sea. The Lincoln Memorial, the World War II Memorial, and the Reflecting Pool all sit on earth reclaimed from Potomac River tidal flats in the early twentieth century. Treasure Island was built in San Francisco Bay for the 1939 Golden Gate International Exposition. Chicago’s Northerly Island was filled in from Lake Michigan to complete Daniel Burnham’s 1909 plan of the city.

In total, around eight percent of the land in America’s major coastal cities was underwater in the 1890s and has since been reclaimed. This includes the land under several major airports, like Newark, Logan, and SFO, as well as neighborhoods like the Financial District in San Francisco, the Back Bay in Boston, and Camden in Philadelphia. Some cities, like Boston and Charleston, have doubled in size by reclaiming land.

Today, reclamation should be more common than ever. Land values in some cities are thirty times what they were in 1950, and high-tide flooding is four to eight times as frequent. Reclamation could extend and protect our coastal cities as it has for centuries. But rather than reclaim more land, we have virtually ceased to reclaim any at all. Since the completion of Battery Park City in 1976, there has not been a single major urban land reclamation project in the United States and only a handful of port expansions.

Subscribe now

Two explanations are commonly proposed for this: first, that the easy spots have already been reclaimed, and second, that improved transportation has made reclamation unnecessary since we can expand cities further inland instead. But neither of these matches the evidence. Other countries still reclaim land at enormous scale, under conditions at least as challenging as those in the United States. The transportation story was once plausible, but not any more. Transportation has stopped improving, and downtown land values have risen so much that reclamation costs would be dwarfed by the value of the land created.

The timing points to a third explanation. Reclamation stopped abruptly in the 1970s when a wave of environmental regulations made it enormously expensive to reshape the landscape. And it halted at the same time in every other country that passed similar laws.

If the legal barriers to reclamation were lifted, we could build hundreds of thousands of new homes near the centers of our most valuable cities. We could build new airports to refresh ailing transportation infrastructure, and we could protect low-lying coastal areas from sea level rise. The disappearance of land reclamation is a choice that we have the power to undo.

Have we run out of good spots?

A common explanation is that we simply ran out of good places to reclaim. According to this view, there used to be plenty of shallow, semi-dry plots available, but they’ve mostly been used up.

There are indeed a finite number of shallow bays near major cities, and since land is often raised in one area by dredging the bottom of the surrounding waters, land reclamation can eat into its own expansion opportunities. But even so, many countries have been adding land for centuries without running out of good options. China has the longest reclamation history of any nation, stretching back to the Tai Lake Basin in the fifth century BC. Low-lying wetlands along the Yangtze river were made suitable for building by using dams and drains. If the US hit the limit of suitable areas for land reclamation within two centuries, China would surely have long since run out, having begun the practice over two thousand years ago.

Subscribe now

But China still leads the world in land reclamation, with close to 80 percent of all land reclaimed in the twenty-first century. Since 2000, Chinese cities of over one million people have reclaimed a land area equivalent to 39 Manhattans, with even more reclamation occurring outside city limits.

American history argues against this theory too: reclamation wasn’t confined to a few choice spots in the most valuable cities. It was common along both coasts, even in smaller cities like Charleston, Wilmington, and Portland.

Hundreds of square miles of water near major American cities are shallow enough to reclaim easily. When Boston’s Back Bay was filled in the late-nineteenth century, it had an average depth of 20 feet. Two thirds of the San Francisco Bay is shallower than this. The South Bay, adjoining Silicon Valley, is less than six feet deep. Outside the main navigation channels, most of New York harbor is less than ten feet deep. Almost all of the water between Miami and Miami Beach, and along the shelf extending south to Key West, is also less than ten feet deep. All of these could be candidates for reclamation.

Nor is depth a hard limit if land values are high enough. Monaco’s recent Mareterra project expanded the country’s land area by three percent by dredging down to 164 feet.

Soil and rock conditions also matter, but much of the American coastline has better conditions than sites where reclamation has already succeeded. Singapore’s Jurong Island is a massive industrial zone created on land reclaimed between 1995 and 2009. The water was more than 50 feet deep and atop another 50 feet of soft marine clay, which settles more than other materials under pressure. To reclaim the land, these layers had to be threaded with long plastic tubes to allow drainage and loaded with a large, heavy quantity of extra soil.

Subscribe to print

It is possible that the United States could eventually exhaust its opportunities for land reclamation. But given how much further other countries have gone and how much suitable coastline remains, that point is likely a long way off.

Has transportation made land reclamation unnecessary?

Another explanation is that land reclamation was replaced by transportation. When walking was the main mode of travel, cities faced tight spatial constraints. Boston began as a tiny peninsula of about two square miles. The city quickly outgrew its isthmus, but land to the south was too far for commuters to walk from. If Bostonians wanted to build close to the city center, they had to expand into the sea.

Other cities faced the same problem even where land was more abundant. In 1890, while Manhattan’s population was similar to today’s, there were still farmhouses beside Central Park, only six and a half miles north of Wall Street. The need to build within walking distance of the city center made northern Manhattan effectively unusable, justifying extensions to the southern tip. Even with plenty of dry land available, the city kept filling in riverbeds and marshland around the island. In all, over a quarter of Manhattan’s land was reclaimed before 1979.

Over time, however, land reclamation became less useful. Better transportation unlocked lots of greenfield land that was cheaper to develop than dredging up new land from the water, while the elevator unlocked brownfield land from the sky.

This abundance of new land within the commuting zone of cities was a cheaper substitute for reclaimed land. Boston’s Back Bay reclamation in the 1860s cost about $400,000 per acre in 2025 dollars. Although the state of Massachusetts ended up selling the land for a healthy profit, by the 1910s there were cheaper ways to unlock land within reach of the city center. If each of the 332 stations of New York’s 1913 subway extension connected a half-mile radius to the downtown, this form of land reclamation cost only $65,000 per acre in 2025 dollars, five to ten times less than land reclamation from the sea.

Subscribe now

But this alternative source of land has proven difficult to extend indefinitely. First, while there may be land that we can access by expanding our cities outwards into virgin countryside, doing so is politically tricky in most countries, not least the United States. Adding more New York suburbs in New Jersey and Connecticut would mean a modern Robert Moses building more expressways through Manhattan.

Second, with enough population and economic growth, even the vast swaths of land accessible by modern transportation can eventually be filled. The edge of New York’s metropolitan area is already 40 miles from Midtown Manhattan at many points, over an hour’s commute. After rapid improvements in the twentieth century, transportation technology has stagnated. Average commute times have increased by a quarter since 1980, partly because cars haven’t grown much faster and partly because highway construction costs at least four times as much per mile as it did in 1960.

Finally, reclamation adds space near city centers, where land is far more valuable than at the suburban fringe. Cities are valuable because they concentrate economic activity, making it easier to find jobs, customers, and suppliers, and because they generate knowledge spillovers making everyone more productive. Better transportation increases their value by putting more workers, suppliers, and customers within reach. Crime and pollution pushed people out of city centers in America after World War II. But since 1990, this trend has reversed, and the value of land in city centers has grown quickly despite continuing suburban expansion.

Today, downtown land values are high enough to justify reclamation costs even in places where suburban land is plentiful. In 1955, the California Division of Water Resources estimated that reclaiming 23,000 acres of land on the west side of the San Francisco Bay would cost about $330,000 per acre in 2025 dollars. Today, the cost of an acre of single-family-zoned land in San Francisco County averages $24 million, more than 70 times the estimated reclamation cost. Even the cheapest county in the Bay Area exceeds $330,000 per acre.

Subscribe to print

Despite the space opened up by cars, trains, and buses, America’s biggest cities face a massive land shortage. Even as the transportation explanation has weakened, reclamation hasn’t returned. Boston’s Back Bay reclamation was profitable when land sold for the equivalent of $40 per square foot. Land in major cities today sells for ten to a hundred times that, and dredging capacity has grown nearly tenfold since 1950. Yet urban reclamation has disappeared even in the most land-constrained cities in the country.

How reclamation went extinct

Neither explanation is convincing, especially where urban land values would plainly justify reclamation. But there is another problem: the explanations cannot easily account for the timing of land reclamation’s disappearance. A technological or geographic explanation would predict a slow and patchy decline, as different regions responded to different pressures at different speeds. But this is not what happened. In the 1970s, reclamation stopped virtually simultaneously everywhere in the country.

Of the 37 cities for which we can find data, every one reclaimed land more slowly after 1970, with many, such as Seattle and Chicago, slowing to zero. A few, including San Diego, Washington, and New York, do show some growth in land area after 1970. But most of this reflects changes in measurement or the completion of projects already underway.

The abruptness of the change shows up clearly in data on the amount of wetland that is lost per year. Most wetland loss is inland and agricultural, not the coastal urban reclamation we’re advocating. Still, the aggregate trend makes it clear that land reclamation ended suddenly around the 1970s. The annual rate of wetland loss in the 1960s was eight times higher than in the 1990s, and twenty times higher than in the 2010s.

In the 1970s, the United States passed a wave of new environmental laws, with perhaps the most far-reaching effects on land use in the country’s history: The National Environmental Policy Act of 1969 (known as NEPA), the Endangered Species Act of 1973, the Clean Water Act of 1972, the Marine Mammal Protection Act of 1972, the Resource Conservation and Recovery Act of 1976, and others. These laws hit the whole country at once, enormously raising the cost of reshaping the landscape, and are the real explanation for the end of American land reclamation.

Still in force, these laws largely regulate the federal government and not private development. But the moment a private project touches federal land, funding, or permits, it must comply with all of them. For land reclamation, NEPA is the most onerous.

NEPA requires covered projects to prepare a detailed report on the environmental impact of the proposed action and its alternatives. While NEPA itself sets no standards for how detailed these reports must be, decades of litigation have turned them into thousand-page behemoths that take years to produce. The environmental assessment for New York City’s congestion pricing plan, for example, is 900 pages long with 3,000 pages of appendices, despite the project requiring zero major construction. When potential harms are identified, agencies must undertake mitigation: buying and restoring 3,000 acres of panther habitat, say, or donating $16 million to conservation banks. In some cases, the project is simply blocked.

Subscribe to print

Land reclamation must always go through the NEPA procedure because the Clean Water Act requires a federal permit for any discharge of dredged material into waterways or wetlands, including the entire coast. Because land reclamation projects do have significant environmental effects, every project must either complete a full environmental impact statement or produce an environmental assessment with costly mitigation measures before receiving a permit. Environmental impact statements filed by the US Army Corps of Engineers, the agency that oversees the permits for land reclamation, take an average of six years to complete, and the mitigation measures often run into the hundreds of millions of dollars.

The Craney Island extension in Norfolk, Virginia is a useful illustration of how these laws complicate land reclamation. Craney Island is a four-square-mile rectangle of dirt and sand surrounded by dikes at the mouth of the Elizabeth river, built solely to store material dredged from Chesapeake Bay navigation channels. In 1997, Congress approved a study of the possibility of expanding the island to the east, extending the lifespan of the disposal area and adding a container shipping terminal to the Norfolk port.

By historical standards, this would have been an easy project: the original 2,500-acre Craney Island was built in just two years between 1956 and 1958. But because the extension involved discharging dredged material into US waters, it needed a Clean Water Act permit. The Army Corps’s environmental impact statement took nine years and ran to 1,400 pages.

Virginia’s environmental quality department then issued permits in 2010, three years after the environmental impact statement was published. After a decade of delays and budget uncertainty, the Corps changed its discharge plans in 2023, which required yet another supplemental environmental assessment. Twenty-eight years after the feasibility study began, almost nothing has been built. The project is expected to finish ‘on schedule’ in 2065.

Craney Island is not an isolated case. Land reclamation ground to a halt across the country after the major environmental laws passed in the 1970s. Very few US reclamation projects started later in the twentieth century, and no notable urban reclamation projects have taken place in the twenty-first.

Evidence for environmental law’s role in the end of American land reclamation can also be found in other countries. It declined abruptly in the 1970s and 1980s across the West. What western nations shared was the rise of powerful environmental movements hostile to reshaping the landscape.

Germany reclaimed over 80,000 acres from the marshy Wadden Sea before 1990. Tens of thousands more were planned, but fierce environmental opposition ended the practice in the late 1980s, turning the Wadden Sea into a national park and making wetlands protected areas. In the 1960s, France used reclamation to develop its Mediterranean coast into a world-class tourist destination, but environmental protection laws in the 1980s banned all construction within 100 meters of the shore.

Subscribe to print

In the early twentieth century, Canada reclaimed tens of thousands of acres of floodplain in British Columbia and the Maritimes, along with hundreds of acres of urban land in cities like Toronto, but large projects like these have stopped or greatly slowed due to environmental rules. In nineteenth-century Australia, hundreds of acres were reclaimed in Sydney Harbor, including the land under the Sydney Opera House, but twentieth-century environmental rules have prevented future expansion.

Even the Dutch, history’s most celebrated reclaimers, largely ceased the practice after the passing of their own environmental laws. From about 1917 to 1970, the Dutch undertook the largest reclamation project in history: the Zuiderzee Works, which expanded the Netherlands by five percent. The fastest phase, the 1927 Wieringermeer Polder, dammed and drained 50,000 acres in just three years. The 1957 Eastern Flevoland Polder then reclaimed nearly three times the area in less than eight years.

But only four of the five planned polders were filled. The fifth, the Markermeer, which was dammed and ready to drain in 1975, was never completed. It would have provided the closest land to Amsterdam, where house prices have risen 120 percent since 2010. In 1969, the Netherlands created a new approval process for large spatial planning projects similar to the NEPA. This process required extensive public consultation and referendums which, while not as explicitly environmental as NEPA, were still dominated by environmental concerns. While the debate dragged on, the 1979 European Birds Directive added another layer of environmental protection. Environmental groups campaigned to halt reclamation permanently and have the unfilled lake designated a Special Protection Area.

In response, the government eventually abandoned its land reclamation plan. In 1992, the Markerwaard was listed as a protection area, preventing any development on some of Europe’s most valuable water. (Some reclamation has resumed since 2014, but not for people: a €300-million, decades-long project is building swampy archipelagos in the lake as bird habitat).

Even the rare successes are instructive. The one major Dutch reclamation of the twenty-first century, an expansion of the port of Rotterdam called Maasvlakte 2, took eighteen years to complete, including an eleven-year planning and environmental compliance period before construction began. Its environmental review alone ran to some six thousand pages.

Reclamation has continued only in countries where environmental regulation is less onerous, like China, Singapore, and Japan. China has reclaimed over 5,000 square kilometers in the first two decades of this century, including a new city of 500,000 outside of Shanghai, and Singapore has expanded its territory by over a quarter since 1975.

Japan does have high environmental standards and a NEPA-style law requiring environmental impact reports. But Japan’s system has two advantages: clear numeric thresholds for when reports are required and courts that make it harder to overturn agency decisions. Japanese infrastructure projects don’t need the defense against legal challenges that make Western projects take decades. The result is abundant housing and infrastructure, including multiple airports, port facilities, and residential areas built on reclaimed land.

Subscribe to print

Land reclamation has always faced obstacles, and the story of its historical fortunes is not simple. But the weight of the evidence points in a clear direction. In America, as across the West, land isn’t reclaimed today because of barriers of geography or technology, but because 1970s environmental regulations made it prohibitively expensive and slow.

The wrong solution for an old problem

Reclamation would be feasible and profitable in the West today if some environmental laws were changed. But those laws were passed to solve real problems. So is it worth bringing land reclamation back?

The postwar state had incredible power to act unilaterally and grade its own work. Robert Moses could demolish entire neighborhoods for expressways without accountability; urban renewal programs replaced them with awful housing blocks; the Cuyahoga River caught fire from wanton industrial pollution; the interstate highway system demolished half a million homes to make way for highways through urban areas; and indeed land reclamation was sometimes responsible for the loss of valuable wetlands. These excesses understandably created fear of unchecked development.

Environmental laws provided necessary curbs, but since the 1970s, the costs of compliance have grown as NEPA reports expanded from ten pages to thousands and spurred half a century of expensive litigation.

Simultaneously, the value of foregone construction has never been higher. Real housing prices are at least five times higher today than in 1970 in New York, San Francisco, Boston, and LA, and three times higher across the country as a whole. Land reclamation may be the most striking casualty of 1970s environmental law. Land reclamation makes housing cheaper, cities more productive, and transportation networks more efficient. It also addresses twenty-first-century environmental challenges: city dwellers produce far less carbon per person than suburbanites, and reclaimed land can make coastal cities more resilient to natural disasters.

Subscribe to print

Despite the environmental costs, there are many places in the country where the benefits of land reclamation are too great to ignore, in particular in and around our biggest cities. In some cases, the environmental benefits of bringing more people into dense, efficient cities where they have far lower emissions per capita may outweigh the loss of aquatic habitat all on its own, leaving the direct benefits of more land as pure reward.

Land reclamation is not the solution to everything, but there’s no good reason we should be worse at land reclamation than we were 150 years ago. There is plenty of land that is easy to reclaim and plenty of areas where the value of land is far greater than the cost of acquiring it. The benefits of more land in our most productive cities are large enough to justify the effort of reforming the laws that currently prevent it. Let’s make more land!

Zigmund Forrest is an independent researcher. Maxwell Tabarrok is a graduate student at Harvard University.

Artwork by Filip Setmanuk.

Chernobyl was the world’s worst ever nuclear disaster. The reactor was built to a flawed design. In 1986, its operators disabled automated safety systems and removed control rods to perform a late-night test. The coolant water overheated, speeding up the reaction until it flash-boiled into steam and violently shattered all the nuclear fuel in a matter of seconds. The graphite in the reactor then caught fire.

Six hundred workers were present when this occurred. Of these, 134 received high doses of radiation, between 800 and 16,000 millisieverts, mostly within hours. Twenty-eight of these died in the first three months, and another 19 died before 2004, though not all from radiation-related causes. While first responders have shown a slight increase in rates of leukemia, there has been no increase in solid cancers.

When the reactor exploded, it released massive amounts of radioactive material into the atmosphere. This ash settled onto the grazing regions of Ukraine. Dairy cows ate it and secreted it into their milk. Every person who drank this milk ended up with some quantity of radioactive iodine concentrated in their thyroid gland. Since children have smaller thyroids, their effective doses were much larger. The hundreds of thousands of children exposed have been closely screened in the years since, and authorities have identified 6,000 thyroid cancers. Fifteen people have already died of these cancers, and 200 might in the long run. These thyroid cancers were completely avoidable because the iodine has a half-life of eight days. After the Windscale fire released radioactive fallout onto grazing land in 1957, Britain discarded contaminated milk for 44 days, thereby avoiding any harm.

Subscribe now

Chernobyl is the only accident in commercial nuclear history that has exposed people to large enough doses of radiation to poison and kill them. But even it has caused only hundreds of early deaths, despite the exposure of millions of people in the exclusion zone and nearby. Radiation impacts on Scandinavia and Germany, where there were major fears about the effects of the fallout, were nugatory. Evacuations and relocations to avoid small additional background radiation levels may have caused more harm than they averted. The same is true of Fukushima and Three Mile Island, the other two large nuclear disasters, but to an even greater extent: neither saw any responders die of the direct effects of radiation, and neither shows any clear impact on cancer rates.

Two years before Chernobyl, an explosion at a pesticide plant in Bhopal, India, released toxic methyl isocyanate gas that killed at least 2,000 people instantly, permanently disabled another 4,000, and caused 550,000 injuries in total. In 1975, the Banqiao Dam in China failed, flooding 12,000 square kilometers, drowning at least 25,000 people, and destroying perhaps five million houses.

Whereas Chernobyl is a household name, Bhopal and Banqiao are mostly familiar only to specialists. People have much greater familiarity with and concern about the risks created by nuclear power, and the world’s international radiation protection regime is based on the idea that any release of radioactive material from a nuclear power plant is intolerable. This has led to regulations that have increased the costs of nuclear electricity over time to the point where it is widely considered a slow, backward, and ineffective technology.

The idea that any release of radioactive material is an intolerable disaster rests on the claim that radiation is harmful even in small, spread-out doses. But this claim is not well supported. The studies that claim to show harms from slow, drawn-out ionizing radiation are unconvincing. By giving them undeserved credence, we may be foreclosing one of the world’s most powerful technologies.

Trouble in Taipei

Between 1982 and 1984, recycled steel contaminated with cobalt-60 was unknowingly used in the construction of more than 180 buildings in Taiwan, including schools, businesses, and approximately 2,000 apartments. Cobalt-60 emits deeply penetrating gamma radiation, which in high doses is fatal: in 2000, three people died from acute radiation poisoning when a Thai scrap merchant took apart a radiotherapy machine containing large amounts of it. Over two decades, more than 10,000 people occupied these buildings, receiving an average total radiation dose of 400 millisieverts, about seven times more than the average American would receive over the same period from background radiation. This makes the Taiwanese buildings a good test of the claim that chronic background radiation at low daily rates damages human health.

A series of academic papers reconstructed the doses that different affected inhabitants experienced, combining thousands of on-site readings with estimates of their time spent at home. CJ Tung and colleagues built a model in 1998 of 24 of the 1,327 apartments that had then been identified as irradiated, breaking them down into one meter by one meter squares, and using dosimeters to measure radiation in those squares, to get an idea of how residents might have been affected.

Doses varied substantially. Most units received trivially low doses, less than the radiation exposure from one full-body medical CT scan per year (ten millisieverts). A small fraction of units received nearly 1,000 millisieverts per year, equivalent to a hundred CT scans. Later studies that mapped out these doses in more detail found that some sections of rooms were exposed to over 8,000 millisieverts per year, equivalent to about two CT scans every day.

A 2004 preliminary report, followed by a 2006 study (Chen et al), was the first to attempt to measure the health effects in this group. There were no cases of radiation sickness nor any major chromosomal aberrations. Affected Taiwanese had biomarkers of cellular stress nearer the upper end of normal reference ranges, but none linked with observable health effects. Cancer rates were, unexpectedly, dramatically lower than in the population at large.

Chen interpreted this as evidence of the health benefits of radiation. This theory, known as hormesis, holds that low doses of stressors, including ionizing radiation, can improve health (in this case, reducing cancer risk) by triggering the body’s repair systems in much the same way that exercise improves fitness by stressing the cardiovascular system. While popular among a small community of researchers, it has not gained widespread acceptance due to limited and conflicting evidence in humans.

This interpretation of the Taiwan data is dubious because it wasn’t able to control for age. Those affected by the cobalt-60 radiation were far younger than the average Taiwanese. Since younger people are less commonly diagnosed with cancer, this is a very significant limitation. A US Environmental Protection Agency (EPA) review, describing the Taiwanese apartment data alongside some similar cases, warned that ‘studies on these populations are ongoing and suffer from various shortcomings, including incomplete follow-up, dosimetric uncertainties, limited statistical power and confounding’.

Torture the data and it will confess

Later studies fixed most of these problems, linking good estimates of the doses individuals experienced with their health outcomes and adding the missing age data. Contrary to the previous report, the studies generally claimed to find that low doses of ionizing radiation can cause cancer. But drawing this conclusion from them requires an unreasonable manipulation of statistics.

A 2006 study (Hwang et al) in the International Journal of Radiation Biology, for example, found slightly elevated incidences of thyroid cancer in women and leukemia in men across a sample of 7,271 people. But overall cancer rates were still lower in this group than in the general Taiwanese population, even after controlling for age.

The way the researchers found these higher rates of cancer was to break cancer cases down into 77 subtypes, such as tongue cancer in men or kidney cancer in women, and see if any of these had higher rates than expected. This approach has two problems. The first is that when you divide cancer cases into 77 buckets, even small and random variations in one or two of the buckets can create the appearance of a relationship. Every statistically significant site-specific result showing higher cancer incidence relies on seven or fewer observed cases. Across 77 different statistical tests, we would expect to see statistically significant results in four cases purely due to random variation. The second problem is that the authors did not register their hypotheses in advance: we have no idea whether they sliced the data into hundreds of different tests before cherrypicking those that gave them statistically significant correlations.

Share

The biggest issue with drawing the inference that low-level radiation harms human health is that the exposed population actually have lower cancer rates than unexposed people. Just like in Chen et al, the number of total cancer cases observed is substantially lower than expected.

Don’t back down, double down

A 2017 study in the British Journal of Cancer (Hsieh et al) follows up on this result, with ten years more data on whether affected people got ill, again using more accurate dose-rate profiles, more closely associated with particular individuals. Like Hwang et al, they claim their findings show that ‘low-dose-rate radiation exposure increases risks of breast cancer and leukaemia’, but the data is similarly brittle.

Within their sample of around 6,000 exposed Taiwanese apartment dwellers and schoolchildren, female breast cancers were higher at higher levels of exposure: 100 millisieverts of exposure was linked to an 11 percent higher cancer risk. This result is statistically significant. The study finds a few other results that are narrowly significant, but only out of a table of dozens, just like the previous study.

But again, the studies fail to state which tests they planned to do beforehand or that they began the analysis with the hypothesis that leukemia (but excluding multiple myeloma and chronic lymphocytic leukemia) would go up after a delay of two years, whereas rectal, stomach, liver, and thyroid cancer would be unaffected – or that cancers overall would be no higher in the study group than among totally unaffected Taiwanese. If they’d entered their research with this hypothesis, it might make the results more credible. Instead, they appear to have stumbled across a handful of correlations by chance.

And once more, the authors don’t highlight that the irradiated group has significantly lower overall cancer rates than the general Taiwanese population.

While we can comb through the data and infer that the average age of the subjects in Hwang was 33. The samples are slightly different, so we can’t just add ten years to get an average age of 43 for Hsieh. The UN Scientific Committee on the Effects of Atomic Radiation estimated that the average age of the Hsieh sample was about 45, seven years older than Taiwan’s population overall in 2012. By 2012, the irradiated apartment dwellers had experienced 247 cancers, across 97,106 person years. Yet the cancer rate for Taiwan as a whole between 2008 and 2012 was about 390 per 100,000 person years.

Overall, cancer rates seem to be almost 35 percent lower in the affected group. If being irradiated with hundreds of millisieverts per year appears to have no effect, or even reduces cancer rates compared with the general population, this strongly implies that the links these two papers identify between much smaller doses of radiation and particular kinds of cancer are random noise.

Subscribe now

Made aware of this, Hsieh responded (as did Hwang) that these gaps must be down to socioeconomic status. Is this true? The best argument provided is that these were residents of new buildings constructed during an economic boom, suggesting that the apartments were relatively expensive. Though no data on the actual socioeconomic status of their residents exists, we can make some educated guesses.

First, though new, some of the buildings were public housing intended for poorer residents or public employees. This makes it unlikely that their inhabitants were wealthier people projected to have better health outcomes than the rest of the population.

Second, even if it were true that the residents were richer than average, this still wouldn’t account for the decreased cancer rates. Existing data suggest that Taiwanese health differences by socioeconomic status do not close this gap: overall cancer incidence is ten percent lower among doctors than in the general population. The incidence of colorectal cancer, one of the most common cancers, is actually lower (by about 20 percent) at the bottom of the wage ladder in Taiwan, though it is higher among those with the lowest education. None of these variations approach the 35 percent lower rates of cancer seen in the irradiated apartment residents.

This is not to claim that Hsieh’s paper provides strong evidence for the health benefits of radiation, as some fans of nuclear power have suggested. Most likely, these results are evidence of nothing.

The perils of low-dose science

From 2011 onwards, scientists began to realize that significant numbers of famous, eye-catching findings were unreplicable, in what became known as the replication crisis. Starting in psychology – once-popular ideas like power posing or the brain-enhancing effects of classical music began to be debunked – before the same occurred across every discipline.

The Taiwan papers have the classic tell of such findings, using tricks to turn random noise into statistically significant results, a phenomenon known as p-hacking. One such trick is dicing a population up into as many different subgroups as possible, expanding a researcher’s likelihood of finding something by chance.

It is inherently difficult to find out if low doses of radiation affect health. Populations vary hugely in all sorts of health, lifestyle, and genetic factors. The anticipated cancer effects of radiation would be low, even for those who expect to find them. Researchers would need a huge sample size and very carefully designed analysis to have any hope of detecting a signal amid the noise. INWORKS, first published in 2015 and updated in 2023, is the most serious attempt. It looks at cumulative low-dose radiation among 300,000 nuclear workers in France, the US, and Britain, measured by badge dosimeters, and finds that each additional 100 millisieverts of lifetime ionizing radiation exposure is associated with five percent higher cancer mortality rates, adding about one percent to the lifetime risk of dying of cancer (compared to an average of about 34 percent).

Subscribe to print

For comparison, 100 millisieverts is about ten full-body CT scans, whereas the median INWORKS worker experienced a four-millisievert dose. Eating 50 grams of processed meat per day is associated, after controls, with a lifetime colorectal cancer risk 18 percent higher, and smoking with a lung cancer risk 1,500 to 3,000 percent higher.

The INWORKS study is the best single piece of evidence for an effect of cumulative low-dose-rate radiation on health, but the size of the effect is small enough that it is far from a given. For one, the INWORKS study does not attempt to account for background doses, despite the fact that different people will have different levels of background exposure, depending on factors like where they live and how many medical scans they have. The average lifetime background dose in the UK is about 70 millisieverts, similar to that in the US and France, meaning that more than 90 percent of INWORKS subjects got most of their exposure from natural background radiation.

Another problem is that INWORKS’s significant result is driven by subjects between the 75th percentile, meaning lifetime exposure of roughly 20 millisieverts, and the 90th, at roughly 50 millisieverts. The relationship between exposure and cancer is not significant for doses between 0 and 20 millisieverts. Only when expanded to doses between 0 and 50 millisieverts does a significant effect appear, before weakening steadily as higher doses are added in, though not to the point of becoming insignificant.

This sensitivity makes the results fragile and is one reason why many look instead at studies with higher doses. The median INWORKS worker received a dose of four millisieverts, which according to INWORKS’s model, would correspond to increasing the lifetime risk of dying from cancer from the US average rate of 17.2 percent to 17.23 percent. An effect this small is hard to confirm even with 300,000 subjects, given so many unmeasured confounders.

Historically, many people have received doses of thousands or even tens of thousands of millisieverts, similar to those of the Taiwanese apartment dwellers. There are cases of Keralans exposed to sand containing thorium; Manhattan Project scientists who inhaled so much plutonium that they peed it out; early British radiologists who worked without any protection from the X-rays they worked with. These groups suffered long-term health effects only if they absorbed high dose rates.

One of the most horrific and notorious cases of radiation exposure was that of women employed to paint radioactive radium onto watches for glow-in-the-dark dials. Early dial painters were told to lick their paintbrushes for greater precision. Those that did ingested radium at extremely high internal doses and developed necrosis of the jaw and often-lethal bone cancers. But the thousands that did not lick their brushes, despite receiving doses that were often hundreds of times higher than normal, did not see any higher rate of cancer.

Though unlikely, it is possible that future studies will reveal a more substantial correlation. In the meantime, the lack of any strong signal in the Taiwan data is one of many good indicators that low-level radiation is not a cause for concern. Where radiation does harm people is when they are exposed to high, acute levels that even the worst nuclear disasters seldom release. When people are looking this hard, absence of evidence is, in fact, pretty good evidence of absence.

Alex Chalmers is a staff writer at the Cosmos Institute and a former editor at Works in Progress.

Ben Southwood is an editor at Works in Progress.

One of the most expensive projects in history is under construction in Abilene, Texas. This joint venture, Stargate, is the flagship of a bigger project by the same name led by OpenAI and Softbank, and is expected to cost well over $40 billion for a high-performance computing campus that will train new generations of AI models.

Stargate is just one major project in one of the biggest investment booms in history, driven by the belief that increasingly powerful AI models can deliver explosive economic growth. But it will require enormous amounts of electricity to work: Stargate is expected to draw 1.2 gigawatts, as much as 313,000 median American family homes, at peak load. A report by EpochAI and an energy research institute projected that total AI computing power would reach 100 gigawatts worldwide in 2030 if the 2025 growth rate stays steady. And data centers aren’t the only energy-hungry element of the AI revolution. The biggest battery manufacturing plants in the US draw energy at a rate of 115 megawatts, and the first phase of TSMC’s Arizona semiconductor plant will draw 200 megawatts.

The primary bottleneck to this growth is the availability of electricity. But this doesn’t mean there is an energy shortage. Instead, the constraint is connecting the flood of new data centers and the plants to power them to the electric grid. Before any new piece of infrastructure can be connected, grid operators must study how it will change power flows around the grid and determine whether upgrades to the system are required. That process is significantly backlogged. Though the median power plant in 2005 waited less than 20 months for interconnection, this had jumped to 55 months by 2023.

The interconnection process wasn’t created for today’s world. Grids use an inflexible first-come, first-served queue that leaves some of the most valuable projects stuck behind less important ones. They also evaluate according to rigid conditions that don’t reward plants for being willing to cover their own power needs for short periods. To prepare for the AI age, grid processes need to change.

A power-hungry future

Estimates vary for how much power will be needed by the data centers and chip manufacturers of the future, but the heads of every major AI company agree that they will need more than they are currently able to get. Jensen Huang, CEO of Nvidia, has said that ‘every data center in the future will be power-limited’. Mark Zuckerberg said Meta ‘would build … bigger [AI training] clusters … if we could get the energy to do it’. And OpenAI CEO Sam Altman told Congress that ‘the abundance of [AI] will be limited by the abundance of energy’.

Regardless of how the data center boom plays out, there is a long-term shift towards electrification across the economy. Electricity can be converted into work instantly, unlike fuels, and with little energy loss. Electricity creates motion directly, while fuels must first be combusted in an engine. This is why electric vehicles can cost half as much to fuel even though electricity is more expensive than gasoline. The simplicity of electric motors also means that electric vehicles have half the lifetime maintenance costs of gas-powered ones.

Electricity also transmits information. Transistors switch on or off depending on the voltage applied to their gates, which allows circuits to perform logical operations. Radios, screens, and computers cannot run on gasoline alone.

Between 1990 and 2024, the price of electric motors declined by 97.5 percent. Power electronics fell 99.5 percent in price, processors built into devices by nearly 99.9 percent, and batteries 98.8 percent. As prices have declined, performance has improved. For example, the amount of energy a battery can store per kilogram has increased five-fold over the same period.

These trends allowed the progression from the Walkman to the iPod and then the iPhone, as well as enabling battery-powered cars, delivery vans, and bicycles. At the same time, the increased computing power built into devices is giving autonomy to newly electrified cars and trucks, robotic vacuum cleaners and mowers, and drones. The commercialization of humanoid robots may also be on the horizon. The future will be largely defined by technologies that run on electricity.

The grid bottleneck

This growth in demand is already straining power grids. Operators are increasingly forced to use expensive local plants because of what grid operators call congestion: cheaper plants are on the other side of transmission bottlenecks and there aren’t enough cables to get the electricity through. In the United States, the additional costs incurred because of grid inefficiencies like congestion ran to $11.5 billion in 2023, an increase of 45 percent from the year before.

ERCOT, the grid that provides 90 percent of Texas’s electricity (including for the plant being built in Abilene), forecasts that it won’t have enough power to meet demand in summer 2028. PJM is the US’s largest grid, both by the amount of electricity it provides and the population in its coverage area, serving an area between Chicago, New Jersey and North Carolina. In 2025, PJM was not able to buy enough future generating capacity to meet projected demand. MISO, another large US grid operating between Louisiana and Minnesota, concluded in one study that ‘resource adequacy risks could grow… absent increased new capacity additions’. PJM’s CEO put it more plainly: ‘We need capacity – a lot of capacity.’

Discussions about expanding electricity supply to power the future often become debates about which source is most suitable: gas, nuclear, solar, or something else. But these are a distraction. Far more fundamental is ensuring power can be efficiently delivered where needed.

When different generation technologies coexist, they can average out to a grid that is cheaper, more reliable, and less polluting than any single type alone. The question isn’t which technology to use, but what balance. But this question isn’t best answered in the marketplace of ideas. Instead, it should be addressed in the marketplace for electricity.

The market for electricity

In the US and Europe, power grids are largely liberalized: provided they can get the necessary permits, independent developers build power plants, and the local utility is required to connect those plants to the grid. In the United States, 88 percent of large-scale power projects currently in development are privately organized and funded.

The power market is run by the grid operator’s economic dispatch software. Each power plant tells the operator its costs, and the operator commissions the cheapest plants, accounting for transmission constraints. As load increases during the day, the operator keeps commissioning the next cheapest plant. The price of power is set at the marginal cost of the next cheapest plant, and that’s the price every power plant is paid.

Market prices signal to power plant developers about levels of supply and demand. In the same way, prices balance different energy sources based on the strengths and weaknesses of each. For instance, as more solar panels are built, the value (and therefore price) of power during the middle of the day, when the sun is shining most, adjusts downward. From December 2020 to September 2025, maximum solar output in ERCOT increased from 4 to 29.8 gigawatts. And from 2020 to 2025, the value of power at 1pm relative to the highest-priced hour decreased from 92.9 percent to 38.7 percent. As one technology type becomes overbuilt, prices reflect that and developers react accordingly.

The evolving daily price shape in response to the abundance of solar energy was a signal that the grid needed storage capacity, and power plant developers responded. From 2020 to October 2025, ERCOT went from having almost no battery storage to a combined battery discharge of 8.6 gigawatts. The same process has played out in California and many European markets.

One might assume that the price of electricity for consumers is dictated by market forces, like those that regulate supply and demand across different power plants. But to a large extent, it is not. Grid infrastructure, like large power lines, is generally planned by the grid operator, and the cost is passed on to consumers at a price approved by state and federal regulators. In one typical utility territory within ERCOT, the portion of regulated costs on the average residential customer’s bill has grown from 28 percent in 2002 to 40 percent in 2025.

Regulators of all major grids have set caps on wholesale prices in response to public outrage at price volatility. As a result, the generators needed to keep the grid reliable are sometimes unprofitable. Under normal market circumstances, generators would stop running when it wasn’t profitable to do so. Instead, regulators put further policies in place to prevent this. Markets for capacity mean that generators are paid not for energy itself but for committing to be available if there is extra demand. Must-run agreements pay unprofitable generators to keep running if they keep the grid reliable. These are negotiated bilaterally between generators and the grid operator, outside of any wider competitive process.

Policy choices also shift the equilibrium. For example, America’s Inflation Reduction Act gives a $30 tax credit for every megawatt-hour produced by qualifying renewables. Power plants that opt in, typically wind, paradoxically often offer their power at negative prices, making money from the tax credit even when they literally pay consumers to use them. At times, a large enough share of the market offers electricity at a negative price that electricity overall (not just from one supplier) can have a negative price. A similar dynamic is playing out in some European countries.

Despite these interventions, which make them less efficient, markets still find an equilibrium. The interaction of supply and demand creates prices that power plant developers use as indicators for what the market does or doesn’t need more of. If prices are high, new power plants enter. That’s why arguing about the best power generation method is overrated. Well-designed energy markets answer this question automatically. The real bottleneck is connecting to the grid at all.

Share

Power struggles

xAI’s Memphis data center operated partially off-grid for months. When it first came online in 2024, it could reportedly draw only eight megawatts from the grid (enough to power a few thousand electric toasters). Rather than wait for its grid connection to be upgraded, xAI installed 422 megawatts of on-site gas turbines. Once transmission upgrades were completed, the project would shift to consuming grid power and the on-site generators would be used only for emergency backup.

Such off-grid generation is a temporary solution. Grid power is more reliable and, on average, cheaper. But thanks to the long queue of projects waiting for connection, 62 percent of data centers are considering off-grid solutions, either to get up and running faster or to improve reliability. Google is even exploring the possibility of a data center in space powered by solar panels. Grids connect to new generators and energy-hungry infrastructure only after studying how to do so using an engineering model that simulates power flows during peak load scenarios. If the new infrastructure would cause overloads on any parts of the system, then those elements need to be upgraded. The utility then needs to estimate the cost of those upgrades and build them.

The grid operators who run these studies are typically public bodies charging a small fee on electricity sales to fund their operations without seeking to make a profit. State utility commissions and the Federal Energy Regulatory Commission have determined that grid operators should run the interconnection queue as a first-come, first-served process that is essentially free.

That process is moving very slowly, creating a huge backlog of mostly phantom projects. Every US grid has more power plant capacity waiting to be connected than there are gigawatts of peak demand. The backlog also exists for the separate but similar queue of infrastructure not yet connected to the grid. Across ERCOT, there were 143.5 gigawatts of data centers seeking to connect as of October 2025. This compares to ERCOT’s highest ever total demand of 85.9 gigawatts in August 2024.

This system was designed for a different era. The last time electricity use grew five percent annually or faster was between the 1950s and 1970s, during the adoption of air conditioning, refrigerators, dishwashers, and washing machines. From 2005 to 2023, electricity use was almost completely flat. But as that lull in growth comes to an end, the grid will have to adapt.

Crossed wires

The main flaw of the interconnection process is that it uses a first-come, first-served queue. This means that high-priority requests can spend years stuck at the back of the line behind other less important ones.

Some requests are speculative, submitted by developers before they have ready customers. Some are duplicative, meaning a developer fishing for a good spot has requested to interconnect the same project in multiple locations. These are among the reasons that 72 percent of requests to connect submitted since 2000 were ultimately withdrawn. Other requests are high quality but small and low value. This creates a feedback loop: the harder it is for developers to estimate the cost to interconnect, and the longer they have to wait for an answer, the more speculative requests are made, further clogging the system.

Every major grid has a roughly similar process. The system impact study comes first. The operator models the grid during peak demand conditions and assesses the cost of any necessary upgrades. After reviewing those costs, the interconnection customer (the developers of the power plant or power-hungry infrastructure) decides whether to proceed. Almost 40 percent of withdrawals between 2020 and 2023 were at this stage. If the customer proceeds, the grid operator then determines how to physically connect the project to the grid. Again, it delivers a quote for these costs to the customer, who then decides if they want to proceed. If so, they sign an interconnection agreement, and the utility begins work on the upgrades.

There are some opportunities to automate these workflows, but each request takes time. The median delay from initial request to the interconnection agreement was 34.2 months for agreements signed in 2023.

Another major issue is restudies. Once a system impact study identifies required network upgrades, those upgrades are added to the grid model that’s used to study subsequent projects in the queue. If those upgrades are not built because the earlier project withdraws, any subsequent studies may need to be redone. In one extreme case, a restudy increased a 242-megawatt wind project’s network upgrade costs from $33.5 million to $99 million after the project was already operating.

The slowness of the interconnection queue adds to households’ energy costs. The clearest example of this is found in PJM, which has the slowest queue of the US grids. In 2025, PJM’s electricity suppliers paid $14.7 billion at auction to make sure they’d have enough power to supply their customers, a huge jump from $2.2 billion the year before. Meanwhile, many gigawatts of renewable power projects have been sitting in the queue for five years or more. If just 30 percent of those projects had been interconnected, the auction’s cost would have been 63 percent lower.

Grids have adopted many sensible reforms to the interconnection queue, such as requiring higher deposits to enter the queue and forcing developers to prove they have land to build on and to disclose if they’re making duplicative requests. The most significant reform has been transitioning to cluster studies, which study multiple projects in the same simulation. This reduces restudies because removing any one project has less impact on the overall solution. Clusters also allocate costs more fairly. This prevents a single project triggering a major transmission upgrade that renders it uneconomical. In cluster studies, the cost of these backbone upgrades are shared among all the projects that would benefit from them. While these were necessary reforms, the backlog has only grown as they fail to address the fundamental irrationality of the interconnection process itself: the inflexible first-come, first-served queue.

Tragedy of the commons

In the early 2000s, regulators wanted to level the playing field between new independent producers and incumbents. They worried that utilities might preference their own projects. As there was plenty of spare capacity at the time, the first-come, first-served system seemed both simple and fair.

But today, transmission capacity is a limited resource and independent power producers are thriving. This means there is much less concern around the power of traditional utilities. In response, many grids are proposing mechanisms to allow more valuable projects to jump the queue. MISO, PJM, and SPP, three large grids in the US, have proposed mechanisms to prioritize projects that are most viable or most necessary for the system’s reliability. But these mechanisms are band-aid solutions.

The flood of requests is a typical ‘tragedy of the commons’. Everyone is incentivized to spam the queue with requests. Auctions can fix this.

In the fishing industry, when tradable fishing quotas were introduced worldwide in the 1970s, what had been a mad dash for fish became an orderly and efficient process. Before the quotas, the whole season’s supply was caught in a few days of dangerous, non-stop fishing. Market share went to whoever bagged the fish fastest. Once quotas were introduced, fishers could time their catches with market demand instead of catching everything at the start of the season and then freezing it. And since more efficient fishermen made more money, they were willing to pay more, and quotas went to them rather than the fastest fishers, lowering prices for consumers.

Auctioning new grid capacity could bring similar benefits. The scramble for interconnection would be replaced with an orderly process in which the highest quality projects would get priority. Developers’ bids would reflect both the likelihood that the project will come online and the projected value of the project to the grid if so. Less viable and less valuable projects would be weeded out.

The simplest way to implement an auction would be to create small ‘fast-track clusters’ that receive expedited studies throughout the year. Projects in the regular annual cluster could bid to enter the fast track, and the highest bidders up to some preset number would be admitted. The fast-track proposals recently adopted by some grid operators follow this structure, except that admittance to the fast-track cluster is based on an administrative scoring mechanism, not developer bids. But administrative scoring mechanisms could never capture the subtleties that a developer’s bid would. For example, developers have private knowledge about a project’s chances to get permitted, like whether the site has trees with endangered bird nests or unhappy neighbors generating pushback.

Only connect

Another inefficiency of the current transmission system is that it’s built for peak load conditions that occur only a few hours per year. The rest of the time, much of the grid’s generation and transmission capacity is sitting idle. In 2024, 42 percent of ERCOT’s capacity went unused for half the time.

Building the network upgrades that allow for this very high peak capacity can take years. But new power plants must wait for that construction to be completed before they can join the grid, even though the equipment being constructed will rarely be necessary. A survey of power developers found that transmission construction was the top cause of project delays.

The solution is to let the power plant come online before the upgrades are constructed on the condition that it agrees to turn off during peaks. The power plant doesn’t contribute to the reliability of the system since it won’t be available when the system is under stress. But it will produce electricity the rest of the time, which will lower costs. This is known as energy-only service.

ERCOT has made this energy-only service, also known as non-firm transmission because it implies no guaranteed ability to export or import power, their default. They call it connect and manage. That’s a large part of why projects move through ERCOT’s queue faster than any other grid operator.

But despite being an option in all organized US grids, 87 percent of interconnection requests by capacity opt for the slower firm transmission service. That’s because non-firm service disqualifies power plants from the payments they could otherwise receive in exchange for being available at times of high demand. Other grids could adopt connect-and-manage if they eliminated this payment system and, like ERCOT, raised price caps to allow energy prices themselves to reflect scarcity value. Short of that, they could make non-firm service more popular by allowing generators to apply for firm service after being interconnected on a non-firm basis.

Subscribe now

Can’t we just use less energy?

These proposed reforms would allow fast interconnection of new power plants. But if AI is a race with national security implications, they won’t be fast enough.

The fastest way to increase capacity for new large loads would be to decrease demand from existing sources. Demand reduction is the option energy planners have always reached for in a pinch. The concept of energy conservation originated in the late 1960s as the US power grid faced resource constraints for the first time in its history. Energy conservation became a national imperative. For the first time, utilities began promoting things like turning off lights and using less air conditioning. Christmas lights were framed as being wasteful and unpatriotic, and in 1973 the National Christmas Tree had only a single light.

Yet while short-term crunches can be fought with conservation, they are difficult to sustain, and increased energy efficiency can simply lead people to using more energy for other purposes. What’s more, the vast majority of electricity users are completely indifferent to the real-time cost to produce electricity, because they’re on fixed rates for terms of months or years. The wholesale cost of a clothes dryer load, which could consume four kilowatt-hours, could range from negative $0.12 to $20, but the retail customer always pays $0.64 at the average US residential rate of $0.16 per kilowatt-hour. This creates irrational behavior, like someone starting a load of laundry while the grid teeters on the edge of a brownout. If customers faced prices more closely linked to real-time costs, they would adjust their energy use accordingly. Prices would be more stable and, on average, lower.

Already, we’ve seen that power costs change throughout the day because of the increasing availability of solar power. Some power suppliers offer time-of-use rates that charge a different rate at each hour of the day to reflect what it costs to provide. As these are adopted more widely, the highs and lows of demand will increasingly follow the same daily pattern as wind and solar availability. Few customers think about their bill enough for small price differences to influence their behavior, but their power-hungriest devices, like their thermostat or EV charger, could schedule themselves to automatically run when the cost is lower.

Daily reconfiguration could be helpful. More helpful would be reducing demand during the few hours per year with extremely high wholesale costs, when the grid is under greatest stress. Even without time-of-use rates, customers can give their utility provider control of their most power-hungry devices, and the utility can turn them down during grid emergencies in exchange for a reimbursement on their bills. The US had 30.5 gigawatts of this capacity, called demand response, in 2022. But the headroom this gives is unreliable because participants may override the utility’s control, turning their air conditioning back up if they are too hot, for instance.

That’s what makes batteries, installed at customers’ homes, the ultimate load-managing tool. Electronic devices continue to work as normal in periods of high demand, just switching from grid to battery power. And batteries offset the demand of the entire home, not just one or two devices. Combined with rooftop solar, they can significantly reduce a home’s reliance on the grid, making room for new large loads. Perhaps the greatest advantage of all is that these at-home power resources don’t have to go through the interconnection queue; the installation process takes weeks to months, not years.

But devices like this free up relatively little grid capacity compared to the huge amounts that will be needed to power the AI revolution. They’re also relatively expensive, as the cost of installation has to be repeated on every home individually. Matching the demand of a single two-gigawatt data center, which could be handled by a handful of utility-scale projects, would require installing solar and battery systems at nearly 175,000 households.

The data center trilemma

Most data centers have on-site generators as backup power for if the grid goes down because they need a stronger guarantee of reliability than the grid alone can provide. But this capacity could have another significant benefit.

Like generators, power-hungry infrastructure could come online on a connect-and-manage basis, interconnecting before all transmission upgrades are complete on the condition that it disconnects from the grid during peaks. Instead of turning off entirely, it would switch to their on-site backup power. And just like connect-and-manage lowers grid costs, flexible data centers would also lower costs for consumers by ensuring less capacity sits idle outside of peak usage.

Because grid peaks are brief and infrequent, data centers would have to disconnect for only a few hours per year. One study found that 76 gigawatts’ worth of new loads could be added (across an area covering most of the US) if these new loads were willing to disconnect during just 22 hours per year. And those 22 hours of disconnection aren’t consecutive. Each disconnection would last a few hours at most, meaning that data centers could rely on batteries rather than thermal generation. The National Renewal Energy Laboratory estimates that grid-scale, four-hour batteries can be built for $1,300 per kilowatt, as compared to around $2,500 for combined cycle gas turbines.

Disconnecting from the grid isn’t exactly ideal, but given the constraints, developers face a trilemma: data centers can be large, they can come online quickly, or they can receive firm grid service, but not all three. If large data centers want to come online quickly, they’ll have to be flexible.

The powerful shall inherit

The data center boom has exposed the weakness of Western grids. But these are not the only new source of demand for electricity that we can expect in the coming years. Industrial electrification, hydrogen production, water desalination, heat pumps and air conditioning to cope with climate change, electric vehicles, and perhaps even electric aircraft are all likely to increase demand.

The twenty-first century will belong to those who make the most of these developing technologies. Without sufficient power, this will be impossible. And the main constraint to supplying power-hungry infrastructure is not power generation – it is getting that power where it is needed.

This is a blessing: insufficient power would be a much harder problem to solve. Instead, the solutions are simple. The queue of infrastructure that needs to connect to the grid should become more flexible, allowing fast-track slots to be auctioned so that the most valuable and viable projects can start sooner. And payment systems should reward both producers and large-scale consumers that are willing to disconnect for a few hours per year when demand is highest. These simple fixes can give America a power grid fit for the future.

Chris Gillett is a market analyst in electric power. His views don’t represent his employer.

Industrial modernity was invented in the English midlands around the end of the eighteenth century, kicking off an unprecedented global rise in living standards that continues to this day. But what made this revolution possible was a lesser-known political revolution that took place in the previous century: a period known in English history as the Glorious Revolution.

Today, the Glorious Revolution is remembered for having introduced the system of constitutional monarchy, which subordinated the power of the monarch to Parliament. What is really interesting, though, is how Parliament used its new power: to untangle excessively strong and complex property rights, making it possible for the first time for sustained investments to take place in infrastructure and agriculture across England. And unlike in most other political revolutions, this happened almost entirely peacefully and with the consent of the people whose rights were being redrawn.

Fragmented political and property rights have again become one of the Western world’s biggest problems, leading to the ‘vetocracy’ that paralyzes many developed countries. Could we foment a Glorious Revolution of our own?

Multiple scleroses

Seventeenth-century Europe was economically stagnant. Despite intellectual advances like the creation of patent systems, the flourishing of mathematical schools, and the founding of learned societies like the German Leopoldina and France’s Académie des Sciences, the physical world saw only plodding progress.

Agriculture was practically the only game in town. Almost everyone lived in small rural settlements: fewer than ten percent of Europeans lived in towns of more than 5,000 people in 1500. By 1700 this had grown to just 12 percent; and by 1800, to 15 percent. In France, Europe’s most populous country, nearly 70 percent of the workforce worked in agriculture between 1695 and 1790. Germany was similar, and Scandinavia, Russia, Eastern Europe, Southern Italy and Spain were possibly more agrarian.

The default was for countries to continue for centuries with virtually no growth. Spanish output per capita was flat for half a millenium between 1300 and 1800. Other areas saw temporary bursts of growth, followed by stagnation or reversion: Swedish and Portuguese incomes were lower in 1800 than in 1550, and the Italian efflorescence during the Renaissance was followed by steady decline for centuries. Even the Dutch golden age of 1500 to 1680, which saw the Netherlands buck the European trend with a large majority of its population working in industry or commerce and living in towns, was followed by more than a century of relative stagnation.

Subscribe now

Property ownership was so fragmented that nobody investing in improvements could expect to make a return. Splintered ownership of farmland discouraged the adoption of new crops and rotations. Rigid inheritance rules made it prohibitively difficult to invest in improvements to land or infrastructure. Roads were bad because nobody took responsibility for them, hindering the transportation of manure for fertilizer and preventing the trade that would allow different areas to specialize in different crops. Agricultural yields were so low that nearly everybody had to work to produce food instead of in other industries like mining or manufacturing.

Many European states tried to solve these problems, and nearly all failed. The reason was usually the same: monarchs depended on the support of landowning aristocrats and clergy, and those landowners had no confidence that reforms would leave them better off.

In 1746, Spain’s King Fernando VI tried to replace his country’s complex tax system with a single income tax to fund public investment, but was blocked by nobles, the church, and municipal oligarchies. In 1776, French Minister of State Anne-Robert-Jacques Turgot tried to liberalize the property system and replace the patchwork of taxes with a single land value tax, but the Parlement of Paris refused to register his reform edict. In 1781, Austrian Emperor Joseph II abolished formal serfdom, but ex-serfs remained effectively bound until 1848, since the crown enforced landlords’ prerogatives.

Only one country succeeded in modernizing its property rights in this period: England. The Glorious Revolution resolved the conflict between landowners and the Crown by handing the country to the landowners. One might expect this to have produced an oligarchy that jealously guarded its privileges. But the landowners, precisely because they were empowered, did something their continental counterparts could not. They dismantled the fossilized property arrangements that had blocked development elsewhere, and in doing so, set the stage for the Industrial Revolution.

The Glorious Revolution

In 1685, James II ascended the English throne. Several previous kings had faced accusations of covert Catholicism, but James II made no secret of his faith. He appointed Catholics to a range of government posts and tried to assert control of local governments, so that he could appoint the officeholders who oversaw elections to pack Parliament with loyal stooges. All this fueled fears that James was imposing Catholic, absolutist rule. According to one contemporary, at least 20,000 opposition pamphlets were circulated across the country during James’s short reign.

In June 1688, seven aristocrats wrote to William of Orange, inviting him to invade England with their support. William was the Dutch Republic’s head of state, married to James’s eldest daughter, Mary and, crucially, a Protestant. In their letter, the seven wrote that ‘the people are … generally dissatisfied with the present conduct of the Government in relation to their religion, liberties, and properties (all which have been greatly invaded)’.

Subscribe to print

William landed at Torbay on 5 November that year with 20,000 Dutch soldiers and English exiles. The English defending army defected or melted away without a fight, and James fled to France, effectively abdicating. William and Mary were crowned as joint monarchs in April 1689.

A new political settlement quickly emerged. Parliament passed a Bill of Rights, reestablishing its own legal supremacy; the Mutiny Act of 1689 forbade a standing army without parliamentary assent; and taxes were decreed to expire after a year, requiring the monarch to call Parliament for renewals. Later acts established regular elections.

Today, this cementing of Parliamentary supremacy is what the Glorious Revolution, as William’s invasion came to be known, is most remembered for. But a second consequence was even more important: dramatically increased legislative productivity.

In the lead up to the Glorious Revolution, James II and his predecessor, his older brother Charles II, had been trying to rule without Parliament, dismissing it before it could pass legislation or simply not calling it. Parliament met infrequently at the end of Charles II’s reign, especially after the Exclusion Crisis of 1679, in which it had sought to pass bills excluding the future James II, by then a known Catholic, from the line of succession. Following this, it did not meet again for nearly all of James II’s short reign.

Before 1688, bills frequently failed because the monarch cut sessions short. After it, Parliament became an extremely active legislature, passing thousands of acts. Before the Glorious Revolution, legislative sessions had been short and variable: the average number of working days per session was 61 between 1660–85 and 57 for 1660–89. While some sessions lasted over 400 days, most lasted fewer than 50. Afterwards, legislative sessions met annually, owing to the requirement to renew military funding and the right to court-martial deserters. They began in autumn or winter and continued for more than 100 days.

Subscribe now

From 1660 to 1689, about 25 acts passed per session. Between 1688 and 1724, the average increased to 50–75 acts per session, rising to over 100 and eventually 200 per session in the late eighteenth century. Access also broadened. Before the Revolution, only the largest landowners brought bills to Parliament. Afterwards, many more minor gentry with no obvious political connections applied. The vast majority of these bills sought to reform landholdings or property rights in some way, either consolidating them, removing legal restrictions on how the land could be used, or allowing developers to charge for access to new or improved infrastructure.

Fragmented ownership

Across early modern Europe, plots of land were small as a result of the traditional open-field system, in which tenant farmers held many small strips of land scattered across different fields, as well as inheritance laws that divided landholdings equally among descendants.

The problems with small fields grew as technology improved, creating the potential for new economies of scale. The horse plough was faster than a team of oxen, but on a farm of many small plots, the ploughman spent more time turning and moving between them, so the advantage was dwarfed by a horse’s greater food cost. Under Russia’s equivalent system, peasants wasted as much as half their time moving between land strips. English farmers faced similar inefficiencies.

Small farms and the open-field system also made it hard to adopt better crops. The medieval three-crop rotation left each field fallow every third year to restore nitrates to the soil. The Norfolk four-course rotation eliminated this dormant period by sowing clover and turnips in the off year, which served equally well to replenish nitrates. But this was almost impossible on shared fields. Turnips and clover were hard to protect from neighbors’ grazing animals, which would not stay within strip borders. And in a fragmented system, anyone who wanted to sow a new crop would have to duplicate the work already shared by those sowing other crops, meaning that switching over required near-universal agreement.

As well as small farms and the open-field system, a third problem was common land: an uncultivated area on which local landowners had the right to graze their animals. The free mingling of livestock made selective breeding difficult, since owners couldn’t control which mated with which. It also meant this land was often overgrazed, since everybody was incentivized to overuse it. Since it was owned collectively, it was difficult to agree to do anything else with it: it could be reorganized or improved only by sending a petition to Parliament showing unanimous support of local landowners. A single holdout could block any change.

Over the century following the Glorious Revolution, Parliament steadily grew more confident in reorganizing land rights, drawing on the legitimacy generated by being elected by, for, and of the landowning classes. It gradually eased the process for enclosing and consolidating landholdings, with petitions eventually needing support only from a supermajority of landowners rather than all of them. By the 1740s, it was passing enclosure laws with 90 percent support of the landowners affected; by the 1750s, 80 percent; and later in the century, 75 percent. The process was formalized in 1801, and the General Inclosure Act of 1845 set up a standing body to enclose remaining fields.

Subscribe now

Enclosure was sometimes acrimonious, since a minority of rightsholders had their property restructured against their will. Some rights to common land were ill-defined, and some people lost them without adequate compensation. On the whole, however, enclosure benefited most people: it made land more productive, and nearly everyone ended up with more valuable holdings than they had before.

Between 1600 and 1815, Parliament passed 3,682 enclosure acts, almost entirely after the Glorious Revolution. This became possible partly because of the government’s increased legitimacy – as government by, for, and of landowners – partly because of the government’s increased legislative productivity – since Parliament sat longer and more predictably – and partly because each reform made sure to have overwhelming local support.

Other countries took centuries longer to solve the fractured land problem. German land was not consolidated until after the Second World War, when millions of acres went through a process like England’s and wheat yields tripled. Taiwan, Korea, and Japan used similar supermajority schemes to consolidate their small fields around the same time as Germany. In early 1900s Russia, Pyotr Stolypin’s land reforms used supermajority petitions by peasant farmers to divide their landholdings, boosting productivity so much that the country became the world’s top grain exporter. The Punjab is the breadbasket of both India and Pakistan in part because both sides consolidated their fields into rational squares, unlike the rest of the subcontinent.

Subscribe to print

Modern agriculture requires consolidated fields, and consolidated fields require someone with the power to reorganize landholdings on a large scale. But before 1800, only the English Parliament, armed with new authority after the Glorious Revolution, managed to carry out a successful program of enclosure.

The dead hand of entail

The Revolution also addressed a second problem: the large quantity of land locked up by strict inheritance laws. European aristocratic families were obsessed with preserving their estates in perpetuity, and horrified by the idea that spendthrift heirs might sell, mismanage, or dismember them. They also wanted to provide for family members outside the direct line without splitting the estate, which would eventually destroy the family.

The details varied, but in England this was achieved through a legal device called a strict settlement. These devices guaranteed payments to widows, younger sons, and daughters marrying out of the family, while preventing an heir from mortgaging, subdividing, or selling any of the estate without their own heir’s consent.

Settlements also included an entail, a long list of successors to whom the estate would be passed. These legal devices are well known in fiction: they disinherit the Bennet sisters in Pride and Prejudice and Lady Mary Crawley in Downton Abbey, and drive the plot of The Quincunx.

Share

Over the 1600s, settlements became increasingly restrictive and harder to break. Conveyancers invented a device that prevented current landholders from undoing the entail. On its own terms, this worked. Large estates tended to be held together. Heirs could not squander their money on gambling and prostitution, and were more likely to marry and have children: owners of entailed estates were nearly 15 percentage points less likely to be childless.

But settlements were inflexible. Landowners could be prevented from selling, mortgaging, or improving their estates without the consent of an heir – even if no heir had yet been born.

This inflexibility could reach absurd extremes. In 1697, Zenobia Hough petitioned the English Parliament for the rights to sell land. Zenobia had a willing buyer, but could not sell her houses, fields, and water mill near Manchester because the settlement gave her husband Daniel the right to manage the property. But Daniel had been missing for 14 years, and no one knew whether he was alive or dead.

An investigative commission in 1829 reported a similar case. A barrister, Hasler Capron, explained that his client had inherited land in 1762 at the age of eleven. The settlement did not include the right to harvest timber. So for 67 years, ‘not a stick [had] been cut’ and ‘all the old trees, with scarcely an exception, [had] become decayed and valueless’.

Subscribe now

This inflexibility left swaths of estates in a bind. An owner might want to mortgage one field to raise the funds to drain others, almost certainly creating a large return. Any conceivable heir would want this too. But if the heirs were underage or not yet born, no mortgage could proceed.

A modern reformer might expect the solution to be simple: abolish entails and grant landowners the freedoms we now take for granted, accepting that many estates would be squandered and tens of thousands of widows and unmarried daughters would lose the stipends promised them.

Parliament, now sitting far more often and able to consider many more proposals than it had before, did nearly the opposite. It heard petitions from both sides, following the longstanding principle of expropriation with compensation, and broke settlements only when the living owner was proposing improvements expected to benefit their heirs as well. Perhaps that is why undoing the property rights of landed elites in this way faced so little opposition, allowing Parliament to pass 3,233 estate acts between 1689 and 1830, each tackling a single estate. These acts clustered around London and the industrializing North West, where the returns from changing land use were highest.

The practical consequences were considerable. In 1695, William, Duke of Bedford wished to build a wet dock on land he owned at Rotherhithe in Surrey, now part of London. His son was dead, which under the old settlement would have prevented any change of use. He and other family members petitioned Parliament, which promptly passed an act vesting the Howland family’s property in trust to raise the necessary funds. The Great Wet Dock (later called the Greenland Dock in the eighteenth century because it was used by whaling ships coming from Greenland) was completed in 1700. It was leased to the South Sea Company in 1725. By 1756 Rotherhithe had three shipbuilding docks; by 1892, eleven, building ships for the East India Company among others.

Estate acts enabled many grand projects. They raised the capital to improve the Bridgewater Canal, often considered England’s first true canal because it did not follow an existing river. The same powers helped finance railways and turnpikes, allowing parts of estates to be sold or repurposed.

Estate acts also helped ‘adventurers’ convert hundreds of thousands of acres of marshy fenland in eastern England into prime farmland. The Grosvenor, Portman, and Bedford families used them to develop Mayfair, Belgravia, and Bloomsbury by selling 99-year leases to developers, long enough to allow them to recoup their costs while eventually falling back into landowner control. Without estate acts, they likely could not have granted leases longer than 21 years.

As with communal land, other European countries failed to solve this problem for decades or centuries. Perhaps a third of the Castile region was entailed by the late-eighteenth century; Spain abolished entails only in 1841. Italy repealed its equivalent in the 1865 Civil Code after unification, and Germany only with the 1919 Weimar Constitution. England’s approach – led by landowners, and taking care to make changes to property rights only when those affected benefited – seemed to lead to much faster reform.

It was possible to fix settlements without landowners’ consent. Napoleon’s Code Civil of 1804 restricted entails on the grounds that they constrained credit, reduced incentives to improve land, and pulled property out of markets. But the top-down approach caused a different problem across the territories where it was imposed: morcellement, the excessive subdivision that entails had been designed to prevent, and the very problem England’s approach avoided.

For whom the road tolls

The Revolution’s third achievement was infrastructure. The division of labor is limited by the extent of the market. Even today, the transport costs of heavy, low-value goods like gravel and sand mean that they are usually produced locally rather than being manufactured at scale in one place and shipped around the country or the world.

This was far worse in early modern Europe, when canals were rare, many rivers were not yet navigable, railways did not exist, and, above all, roads were terrible. In sixteenth-century Spain, wheeled transport was virtually useless; it was faster to travel across open country than to use a road. In late Ancien Régime France, roads were routinely obstructed by quagmires.

Subscribe now

In 1700, English road speeds, which were similar to those on the continent, averaged 2.3 miles per hour in summer and 1.5 in winter. In 1705, the journey from London to Birmingham, a distance of about a hundred miles, took 65 hours. As the historian Cyril Hughes Hartmann put it in 1927, Britain’s roads had been better in Roman times.

The underlying problem was incentives. In the Holy Roman Empire, local governments could charge tolls on roads and rivers. But this led to overcharging without investment. Each stretch had a different authority, and each charged as much as possible while feeling little incentive to improve its own section, since no single section made much difference to the journey as a whole. Timber merchants on the Rhine between Andernach and Dordrecht complained that they had to pay around twenty separate tolls. France and Spain also had dense networks of internal customs gates, including around Paris.

England’s rulers forbade local governments from charging tolls. But this left them with no motivation or ability to improve roads. In both cases, splitting infrastructure into the wrong pieces gave operators the wrong incentives.

Parliament’s solution, after the Glorious Revolution, was the turnpike act. A group of landowners and merchants would form a non-profit corporation, adopt a road, improve it with paving, bridges, and embankments, and charge tolls to cover costs. (Though they could not make profits, better roads increased property income, often by as much as a fifth per acre.) These turnpike trusts usually took over existing roads, though they sometimes added shortcuts and bypasses on new alignments secured through compulsory purchase (eminent domain).

Turnpikes dramatically improved road quality. Muddy tracks were covered with McAdam surfacing (later macadam, and tarmac when combined with tar) or Telford paving. Smoother surfaces meant carriages could be lighter, faster, and easier to pull, needing fewer horses and less fodder.

By 1840, Britain had 20,000 miles of turnpikes, more than twice France’s density of paved road per square mile, and nearly ten times Spain’s. By 1760, average travel speeds in France were half Britain’s. By 1820, British speeds had quadrupled to eight miles per hour, dazzling foreign visitors. Karl Philipp Moritz, a German traveller in England in 1782, called the country’s roads ‘incomparable’, writing: ‘I am astonished how they have got them so firm and solid … A thousand charming spots, and beautiful landscapes, on which my eye would long have dwelt with rapture, were now rapidly passed with the speed of an arrow.’ César de Saussure, a Swiss visitor to England in the 1720s, wrote that the English high roads were ‘magnificent, being wide, smooth, and well kept’.

Road freight costs fell from around 1.2 shillings per ton-mile in 1700 to 0.7 (in 1700 prices) by 1800. Diary surveys confirm that travelers greatly preferred turnpike roads for the smoother, more comfortable journeys they afforded.

A similar model improved rivers and canals. In 1700, freight on navigable rivers like the Thames and Severn cost just 0.12 shillings per ton-mile, compared with 1.2 shillings by road. The problem was that many rivers were shallow and impassable. River improvement trusts dredged channels to open them up to freight traffic, covering their costs through tolls. The total navigable length of British rivers rose from 850 to 1,600 miles by 1750.

Canals were different: they were dug across virgin land and financed by for-profit companies. Four thousand miles of canal were built before 1840, connecting river networks across watersheds and linking industrial towns like Birmingham and Manchester, which lacked natural waterways. By 1850, England and Wales had more navigable waterways per square mile than France, Spain, and Germany.

Together, improvements in roads, rivers, and canals cut transport costs by 75 percent. This enabled a degree of market specialization unseen in the rest of Europe. Manure could now be shipped from where it was plentiful to where it was scarce, on its own raising English wheat yields by a fifth between 1700 and 1840. The Bridgewater canals made Manchester’s industry possible by cutting coal costs there by half.

Regions specialized. Cheshire became a specialist in cheese. Stoke made the nation’s pottery, Burton its beer, and Sheffield its steel. A national wheat market formed, with prices converging across the country. In the 1650s, entrepreneurs in inland Lancashire built a cottage industry, exporting cotton textiles to London and nearby villages. By 1750, and especially by 1800, industries could scale and agglomerate. By 1800, Britain had 95 percent of the world’s cotton spindles, and output had grown more than thirtyfold since the seventeenth century.

Subscribe to print

Across the rest of Europe, internal barriers took far longer to dismantle. German river tolls were scrapped by the 1831 Convention of Mainz; other barriers fell with the Zollverein after 1834. Spanish internal customs disappeared with the Mon-Santillán reforms of 1843–45. England solved its tolling problem earlier and more thoroughly by bringing the immediate beneficiaries into government and letting them drive reform.

Gloriouser and gloriouser

One reason the Industrial Revolution happened in England and not France, the Netherlands, or Germany was that its land, infrastructure, and capital markets had already been transformed.

In 1650, it was difficult to raise capital for a factory, or to transport raw materials from mine to factory and finished goods to customers. There was not nearly enough food to sustain the workers to procure these materials at scale. If the steam engine, the gas lamp, and the spinning jenny had been conceived at this time, it would have been far harder for their inventors to manufacture and sell them.

The Glorious Revolution broke these constraints. Agricultural yields rose by roughly 45 percent, while the share of the workforce in agriculture fell from 50 percent in 1600 to 30 percent in 1800. This agricultural revolution, which came as much from efficient institutions as from new technologies, transformed England’s economy, generating surpluses that fed a growing industrial workforce.

A century after William of Orange landed at Torbay, the England of the late eighteenth century was ready for an incredible acceleration of innovation across every sector, from watches to navigation, that would make it the richest country in the world. Had England remained as institutionally sclerotic as the rest of Europe, the Industrial Revolution could not have been as transformative.

A new Glorious Revolution?

Early modern Europe’s property rights held back growth not because they were too weak but because they were too strong and too rigid. Even those who stood to benefit had no power to alter them. Elites across Europe blocked reforms that would have benefited their nations as a whole.

Similar forces are at work today. Incumbents have property and a stake in elaborate regulatory systems that exist, sometimes openly, sometimes quietly, to protect them and their rights. Zoning, historic preservation, judicial review, environmental regulation all serve, among other things, to protect the property values of existing homeowners in the area. European workers hold what amounts to a right to generous compensation in the case of dismissal, which in practice discourages the restructuring that drives innovation.

Subscribe now

As in premodern Europe, the costs of these tangled rights are high enough that many incumbents would actually gain from well-designed reform. In Palo Alto, every homeowner could be paid tens of millions of dollars to accept Manhattan-level density, and developers might still clear hundreds of billions in profit. Danish workers receive extremely generous state unemployment insurance, up to 90 percent of income for two years, in exchange for giving up labor protections, a tradeoff the state can afford thanks to higher growth.

The lesson from the Glorious Revolution is that reorganizing rights is much easier if reformers work with incumbents rather than attempt to override them. A legislature made up of landowners, which sought supermajority approval for change or took care to make sure nearly everyone benefited from major changes, turned out to be far better at radical reform than any absolutist monarch.

These methods still work. Take housing. Small minorities often oppose any change. But building a coalition for development is as possible today as it was for enclosure. Japan, Taiwan, and Korea each used supermajorities of landowners to replan roughly a third of their cities at higher densities with better infrastructure. A very similar scheme, Pinui Binui, delivers many of Israel’s new homes. In the UK, social housing tenants have approved dozens of proposals to demolish and upgrade their own social housing.

While many of today’s challenges are new, a surprising number have been with us for centuries. But so have their solutions.

Kara Dimitruk is Assistant Professor at Hamilton College and Research Fellow at Stellenbosch University.

Ben Southwood is an editor at Works in Progress.

The Coyote Getter, invented in the 1930s, was a hunting device that spewed cyanide into the mouth of any coyote that bit its scented wool bait. Thirty years later, researchers revisited the device. But rather than revamping it to kill more efficiently, they wanted to replace the cyanide payload with something wholly different: a coyote vaccine.

Wild animal vaccines are remarkable for many reasons, but the most remarkable might be that they exist at all. Vaccines are usually tools civilization reserves to defend the humans and domesticated animals within its bounds. The wild, by contrast, is a place from which we extract resources or defend ourselves, or where we simply leave things untouched.

Subscribe to print

But defense doesn’t have to be adversarial. Around 60 percent of infectious diseases are zoonotic in origin, meaning they are transmitted from animals to humans, and the majority of them originate from wildlife rather than livestock or pets. Ebola came from fruit bats; HIV from chimpanzees; SARS from horseshoe bats and civets; Lyme disease lurks in deer and mice. In theory, preventing disease at its source in wild populations could protect people better than simply reacting to spillovers when they occur.

Preventing human illness is not the only reason to vaccinate wild animals. Many recent programs have been motivated by conservation, such as efforts to protect bats from white-nose syndrome or koalas from chlamydia. There is also budding recognition that these programs are valuable on welfare grounds, sparing animals from pain and suffering.

Nevertheless, wild animal vaccination first emerged as a form of defense. In the mid-twentieth century, laws mandating the vaccination of dogs pushed down the number of rabies cases in the United States, from 74 reported human deaths across the country in 1921 to single digits per year by the 1960s. The remaining cases occurred overwhelmingly through contact with wildlife: bats, foxes, skunks, and raccoons. But worryingly, the prevalence of rabies in wild animals seemed to be growing, both in the United States and in Europe. To keep rabies contained, researchers had to figure out how to immunize these animals, too.

Sick as a dog

Rabies is a terrifying disease. It is transmitted to humans through bites or scratches from an infected animal, such as a dog, bat, or coyote. The virus travels silently through the peripheral nervous system, taking a month or two to creep up into the brain and spinal cord. Once it infects the central nervous system, an array of alarming symptoms emerge. The ancient Greek physician Hippocrates is thought to have been describing rabies when he wrote of ‘persons in a frenzy [who] drink very little, are disturbed and frightened, tremble at the least noise, or are seized by convulsions’. And when these appear, death is virtually guaranteed within days.

Since rabies doesn’t spread between humans, infecting us only incidentally, it’s far more cost-effective to wipe out the disease reservoir in animals than to vaccinate humans. So developed countries have long implemented public health programs to control its spread from animals to humans: vaccinating pets and regulating their movement across borders, neutering or euthanizing strays, surveilling and investigating cases, and rapidly administering treatment after potential exposure. These early efforts worked. By the 1960s, the US had largely brought dog-mediated rabies under control, reducing the number of cases affecting humans almost twentyfold. In 2007, the canine variant of the rabies virus was officially declared eradicated from the United States, with the last recorded case occurring in 2004. Many European countries had achieved this milestone earlier. The United Kingdom had been free of canine rabies since 1922; France, 1960; and Italy, 1973. Human deaths from rabies are now almost unheard of in rich countries, though it continues to kill tens of thousands of people per year in the developing world.

Share

But as canine rabies subsided in the West, another problem arose: wildlife rabies. Between 1938 and 1958, the annual number of rabies cases reported in wild animals grew from 44 to 2,075. Part of the reason for this was increased detection, but there also seemed to be a genuine rise in the number of cases. Rabies was spreading in raccoons, skunks, and foxes across the Americas; in Europe, too, a growing number of rabies cases were identified in red foxes, beginning in Eastern Europe in the 1940s and expanding westward. Genetic analysis of rabies strains indicates that outbreaks on both sides of the Atlantic occurred independently of one another. In both cases, the virus is thought to have jumped from dogs into wildlife.

These infected animals endanger any person who comes into contact with them. Every year, around 100,000 Americans are injected with a precautionary course of post-exposure rabies treatment after wild animal bites or scratches. This treatment, first developed in the late-nineteenth century, is thankfully highly effective, but expensive. Post-exposure treatment also works only if administered before the onset of symptoms, and the prospect of needing it is terrifying. The better solution is to control the disease at its source, in the wild.

Subscribe now

By the 1960s, crude methods like culling and sterilizing rabies-carrying wild animal species had shown limited success, despite hundreds of thousands of foxes, coyotes and other animals being killed. These efforts may even have been counterproductive, as they increase contact between humans and animals, and could spread disease between animals in different areas. With elimination impracticable, prevention was the only solution. In 1962, early research into wildlife vaccination began, led by veterinarian and virologist George Baer at the CDC (at the time the Communicable Disease Center; now the Center for Disease Control).

Animal vaccines work in the same way as many human vaccines: killed or attenuated (weakened) pathogens trigger immune responses, teaching the immune system how to recognize a real infection. What’s different, and more difficult, is delivering those vaccines. Foxes don’t check themselves into clinics, and attempting to capture thousands of feral animals living over vast swaths of wilderness would be a logistical nightmare. Delivery methods would have to be designed differently.

Share

Throughout the 1960s, Baer and his colleagues attempted to solve the distribution problem with imaginative gadgets. The Coyote Getter, as mentioned, was modified to shoot out inactivated rabies vaccine instead of poison. But in field trials, the forceful device often injured animals’ mouths and was unreliable at producing immunity. Another early prototype was CDC veterinarian William Winkler’s Vac-Trap, a spring-loaded needle trap that injected a syringe full of vaccine when stepped on. However, it was expensive, inaccurate, and hazardous. Any animal, or unlucky hiker, could trigger the needle.

The breakthrough came when Baer’s group pivoted to oral vaccination. The rabies vaccine variant used in human post-exposure treatment, where the virus is fully inactivated, must be injected: otherwise, it would be destroyed by the digestive system. But in 1971, Baer’s team discovered that pipetting an attenuated rabies virus vaccine, where the virus has been rendered harmless without being killed, into the mouths of foxes produces immunity. The next step was to package the oral vaccine into something an animal would want to chew.

This concept – live, attenuated virus tucked into a tasty bait – was first field-trialed in Europe. Fox rabies had been advancing westwards across the continent since the 1940s, reaching Switzerland by March 1967. In response, Swiss veterinarian Franz Steck and colleagues built on Baer’s work and began testing a range of vaccine baits, from sausages to dog biscuits to eggs, in the mid-1970s.

Eggs were rejected, as foxes would often save them for later. Sausages and dog biscuits fared better, but the ultimate winner was chicken heads. On 17 October 1978, Steck’s team scattered over 4,000 vaccine-laden chicken heads along the eastern shore of Lake Geneva, creating a barrier against the advancing epizootic. The heads also contained tetracycline, an antibiotic that incorporates into bone and teeth and glows fluorescent yellow under ultraviolet light, allowing researchers to track which foxes had consumed the bait. Within a few years, fox rabies was cleared from the area.

Subscribe now

Following this success, Switzerland scaled up its efforts dramatically. Millions of vaccine baits were distributed nationwide over the next two decades, with individual chicken heads flung onto roadsides and airdropped over the Swiss Alps by helicopter. By 1999, fox-mediated rabies was eliminated from Switzerland entirely.

In the meantime, the US was hit by multiple rabies crises. A raccoon rabies variant had emerged near Virginia in 1977, and spread across the Eastern seaboard over the following decades. To the south, two large epizootics infecting coyotes and gray foxes had spread along the Texas–Mexico border, beginning in 1988. Hundreds of canine rabies cases popped up, and, given rabies’s near-certain fatality, thousands of people faced with potential exposure needed preventive treatment at a cost of around $3,500 per person. Alarmingly, the disease was spreading north towards San Antonio and other major cities. By 1994, the animal outbreaks had become severe enough that Texas Governor Ann Richards formally declared a state health emergency.

Several affected states launched oral rabies vaccination (ORV) programs in response, though with a less grisly bait than deployed in Switzerland. Rather than chicken heads, researchers employed fishmeal blocks covered in fish oil or vanilla. Despite their manufactured, brick-like appearance, these baits proved attractive to hungry animals.

The programs were remarkably successful. The raccoon ORV program halted the westward spread of raccoon rabies, creating a barrier at the Appalachian Mountains that has held for decades. In Texas, every dollar spent on the coyote ORV program saved an estimated $4 to $13 in human treatment – not to mention reducing the risk to human life. By 2008, the state had eliminated the dog-coyote rabies variant. By 2009, the gray fox variant followed. Today, the USDA National Rabies Management Program distributes eight to ten million vaccine baits annually across the US, placed next to dumpsters or in other places raccoons and foxes are likely to frequent, or dropped over large rural areas by plane. The rabies vaccine demonstrated that wild animals could be immunized at scale.

Shielding nature

As the conservation movement grew, the technology found a second application: preserving biodiversity. Wild species have always been threatened by disease, but human activity has exacerbated this. Habitat destruction and climate change have increased crowding in animal habitats and put stress on animals’ immune systems, amping up disease transmission. Humans and livestock also spread disease directly to wildlife, a phenomenon known as spillback.

Subscribe now

The scale of destruction can be enormous. Chytridiomycosis, a fungal disease that infects amphibians’ skin and often fatally disrupts their vital functions, has affected at least 501 amphibian species over the past fifty years, with 90 species going extinct and another 124 declining in population by over 90 percent. In 2015, 200,000 saiga antelopes – 60 percent of the global saiga antelope population – were killed in three weeks by an outbreak of hemorrhagic septicemia, a bacterial infection that triggers sudden internal bleeding. White-nose syndrome is caused by a fungus that invades bats’ skin and disrupts their hibernation, often leading to fatal starvation and dehydration. It has killed over six million North American bats since 2007, with three bat species declining by over 90 percent. Outbreaks like these illustrate how diseases of wildlife, amplified by human activity, can push species toward extinction faster than they can recover or adapt.

Wild animal vaccines are emerging as a promising tool to fight this devastation. In September 2025, Australia approved the world’s first chlamydia vaccine for koalas. While plenty painful and embarrassing for humans, chlamydia is a far more serious threat to koalas, with infections much more likely to cause infertility and often resulting in blindness and death. Disease outbreaks account for upwards of half of all deaths in koalas and are pushing some colonies towards extinction. Alarmed by the threat posed to Australia’s iconic marsupials, Peter Timms, a microbiologist and chlamydia expert at the University of the Sunshine Coast, has spent over a decade leading development of the vaccine.

Distribution, as always, poses a challenge. Since koalas exclusively eat eucalyptus leaves, bait vaccines, like those used for rabies, don’t work. It is feasible, though, to capture koalas. Simply surround a tree with cages, and sooner or later, they will climb down and wander into the harmless traps.

After completing the largest and longest ever study conducted on wild koalas, Timms’s team found that the vaccine reduced the likelihood of developing chlamydia. With vaccines now available for wildlife hospitals, veterinary clinics, and field use, there’s new hope for preventing koala populations from vanishing.

Vaccines for other diseases are still in the pipeline, including vaccines for bats threatened by white-nose syndrome, which wakes them up repeatedly during winter hibernation, draining their stored energy and water until they die from starvation or dehydration. The fungal disease has already killed millions of North American bats, but the stakes extend even beyond their extinction. Many of these bat species are insectivores; without their voracious consumption of agricultural pests and mosquitoes, crop yields could fall and insect-borne disease transmission could rise. For reasons like these, the USGS estimates that bats contribute billions of dollars to the US agricultural economy annually.

Subscribe to print

But their insect-based diet also makes vaccine distribution complicated. Bats, like koalas, aren’t tempted by food baits. But they won’t leisurely wander into cages either. So researchers are experimenting with novel delivery methods instead, such as a vaccine gel or paste that could be applied onto roosting bats, which they would spread as they groom themselves and other colony members. Since September 2025, epidemiologists have been actively conducting field trials at a key site in Wyoming, racing to protect bat populations before the disease wipes them out.

Human kind

From airdropped chicken heads to anti-fungal bat gel, scientists have used ingenious methods to adapt human vaccines for animal use. Some of these efforts have already brought impressive successes, protecting humans and preserving animal species. Peculiarly, in all cases, animal welfare itself is considered incidental – a positive side effect. Even with conservation-driven programs, the concern is for populations, not individuals. A species can be considered safe while millions of its members suffer painful diseases and die, as long as the population remains stable.

There’s a growing group of researchers and activists who seek to go further. Disease isn’t bad solely because it threatens a species’s survival but because it causes suffering. Every koala or raccoon that contracts a fatal illness spends days or weeks vomiting, seizing, blind, feverish, or paralyzed. None can hope for treatment. Then, alone and in pain, they die.

Subscribe now

Wild animal vaccines could mitigate such suffering. Vaccines like the oral rabies bait show that it’s possible to effectively improve wild animal welfare, and there are other painful diseases we could prevent, too.

In 1936, the Coyote Getter was invented to kill coyotes with cyanide. Thirty years later, it was retrofitted to administer medicine. The same ingenuity that made us effective predators could also make us effective protectors. Perhaps wild animal vaccines hint at what we’ll ultimately choose to become.

Michelle Ma is chief of staff at a startup.

Illustration by Jenna Barton.

Correction: This article previously stated that a koala chlamydia vaccine decreased mortality by around 65 percent. The cited study was found to have data handling issues that rendered the mortality reduction no longer statistically significant.

In a corner of the prestigious Kits Point suburb in Vancouver, eleven great towers are beginning to rise from their foundations. This is Senakw (pronounced sen-AHK), a new development owned, managed and championed by the Squamish Nation, an indigenous group native to British Columbia. It will ultimately house 9,000 people, and represents seven percent of Vancouver’s entire projected housing output up to 2033.

Senakw has an unusual history. The land it is built on was home to the Squamish people until they were forced out in 1913. Almost a century later, a court case restored the land to the descendants of those who were expelled, along with almost 100 million Canadian dollars in compensation. Freed from the restrictive planning rules that hold back densification in the rest of Vancouver, the Squamish decided in 2019 to use the land to build apartment blocks that, as well as housing Squamish people, are expected to generate around C$10 billion in income, equivalent to more than two million per person.

This is a story of a dispossessed group that is making the most of a rare and remarkable set of circumstances to generate wealth through upzoning. But it is also a story with relevance all over the world. Senakw shows that when local residents stand to benefit from development, obstacles to housebuilding can be overcome, to the benefit of a much broader group of people.

The village that was lost

The Squamish are one of the Coast Salish peoples, the precolonial inhabitants of British Columbia. Today there are about 4,000 Squamish, and historically, the population inhabited a considerable territory, including the land on which the city of Vancouver was built.

Subscribe now

Wilson Williams, council chairperson of the Squamish Nation, explains that in the precolonial period, the False Creek inlet served as a seasonal gathering place: ‘Senakw was a place of abundance ... It was the mainstay for spring and summer for our people. It was a welcome place for our families during those warm seasons because of the abundance of food’. Salmon, oysters, and berries, key ingredients in Coast Salish cuisine, thrived there. By the 1860s, records suggest it had evolved into a permanent village, featuring traditional Squamish longhouses.

Between 1869 and 1877, eighty acres were allocated by the Canadian government as False Creek Indian Reserve No. 6, more commonly known as the Kitsilano Indian Reserve. Like all reserves, the land was owned by the federal government and held in trust for its indigenous residents, a legal arrangement that would prove crucial more than a century later.

European settlers quickly drove out the existing villagers. The city of Vancouver was incorporated in 1886 and expanded rapidly. Conflict between settlers and natives over space began to worsen. A 1911 amendment to the Indian Act enabled the forced removal of indigenous people from urban areas. Vancouver did not wait: in 1913, the residents of the Kitsilano Indian Reserve village were summoned to meet British Columbia officials under attorney general William John Bowser. They were told they either had to accept a small cash offer for the land or receive nothing. They had no federal representatives to support them, which was unlawful even at the time, and many did not even speak English. Believing they had no choice but to accept, the residents disinterred the remains of their ancestors. Most sailed north to relatives in Howe Sound, and the village was burned.

Contemporary newspaper articles imply that the motive for removal was not primarily financial. At the time, the land was not nearly as valuable as it is now: it sits on the opposite side of the False Creek inlet from the city, and the Burrard Street Bridge that links it today would not be constructed for another twenty years. Instead, newspapers focused on concerns about noise from traditional ceremonies, and alleged immorality and public disorder.

City officials originally intended to designate the land for either courthouses or a public park, but had no fixed plans. The land seizure itself was the primary goal.

After the Squamish residents were evicted, the site was neglected for decades. It ended up being occupied by squatters, who made a living fishing and pulling in logs that floated down the creek. Throughout the 1920s and 1930s, newspapers complained of worsening disorder and mess, as well as the city’s repeated failed attempts to evict the squatters. Ironically, the removal of the original residents in the interest of public order had failed even on its own terms.

Subscribe to print

Throughout the twentieth century, the land was parceled up and passed between various public bodies, including the harbor authority, as well as private companies like sawmills and breweries. From the 1940s, it was used by the Royal Canadian Airforce as a storage depot. In 1967, it was eventually turned into Vanier Park and became a museum complex.

The long fight back

The 1911 Indian Act amendment made it easier to expropriate land from indigenous people. Unfamiliar with legal processes and sometimes lacking a working command of the English language, they found the deck stacked against them. But even in 1913, the province’s actions in the Kitsilano Indian Reserve village were not lawful: the land was federally owned, and the province could not acquire it without federal consent. The Crown immediately began efforts to reclaim the land.

The Squamish people also began organizing. Between 1913 and 1923, sixteen local families amalgamated into the Squamish Nation band. But for the first few decades, they were legally prohibited from pursuing their claim to the land.

Wilson Williams’s great-grandfather, Andy Paull, was a young man when the village burned. Selected at a community meeting to learn English and represent the nation’s interests, Paull studied in a Catholic residential school in North Vancouver and was educated as a lawyer. Williams recalls his grandfather telling him that the seizure of the village was the catalyst for the nation’s focus on strengthening their legal capabilities. But indigenous people were legally prohibited from joining the bar unless they gave up their indigenous status, which Paull was unwilling to do. ‘He became an off-the-record lawyer’, Williams explains, helping not just the Squamish Nation but other indigenous people across Canada.

Share

The Indian Act heavily restricted the ability of indigenous people to seek legal redress over land. Throughout his career, Paull focused instead on advocacy, using his legal skills indirectly to push for the right to launch land claims, a right that was eventually won in 1973.

The eventual breakthrough for the Squamish came from an unexpected direction: academic research. In 1971, anthropologists Dorothy Kennedy and Randy Bouchard began what at first seemed an uncontroversial project: recording historical place names to preserve indigenous peoples’ languages and history. Working with Squamish chief Louie Miranda, they had no idea they were assembling evidence for a legal case that wouldn’t be heard for decades.

Chief Miranda, who had himself been a resident of the Kitsilano reserve, was the Squamish Keeper of the Names, responsible for memorizing ancestral names and history going back generations. Without him, there would have been little to work with.

Following Catholic missionary efforts from the 1840s, almost all Squamish people had converted to Catholicism by 1913. Accordingly, their births, deaths and marriages were usually recorded by the Church. However, Kennedy recalls that ‘the church was reluctant to hand over information’. They did eventually gain access, and not only did they accurately reflect Chief Miranda’s memory, the records also enabled them to ‘bring back hundreds and hundreds of names’, something of great cultural significance in Squamish tradition.

By cross-referencing Miranda’s extraordinary memory against Catholic church records, government surveys, and early anthropological notes, Kennedy and Bouchard could reconstruct not just genealogies but precise information on who lived where. ‘We transcribed all the Catholic records for the Squamish back to 1860’, Kennedy explains. Thanks to the connection between anglicized names in the church record and traditional Squamish names, ‘We could reconstruct where the houses were, who lived there, and their descendants... All of the success was thanks to Chief Miranda’. This painstaking work would prove invaluable in the court battles that began in late 1977, led by Chief Joe Mathias. It took several decades of legal dispute, but the case that turned the tide was decided in 2001: Mathias vs Canada. The case centered on the fiduciary duty of the federal government towards Indian Bands. Several other indigenous groups also made competing claims over the same land.

Subscribe now

One focus was proving that the 1869 residents, when the land was originally set aside as a reserve, were Squamish. The court accepted the anthropological evidence as overwhelming proof, largely because of Kennedy, Bouchard, and Miranda’s meticulous work decades earlier.

Most of the other claims were dismissed by the court, partly due to the statute of limitations expiring. But in order to close the decades of continued legal battles, the federal government came to an out-of-court settlement with the Squamish Nation: it paid C$92.5 million (USD $66.9m) in compensation, which, under the terms of the settlement, would be held in a trust fund for band members. The settlement also included something that received much less media attention at the time, but which has since proven far more significant: 11.7 acres of the original 80 acres of land.

In a ceremony in 2002, the Squamish Nation sailed 28 canoes to the beach in Kitsilano, a symbolic homecoming for the Squamish people. The site did not seem especially valuable: a thin patch of land, approximately half of which lies directly beneath the Burrard Street Bridge in an unusual tri-point shape. With the neighborhood zoned largely for single-family housing, observers did not initially notice its development potential.

The grand bargain

Vancouver is one of the most expensive housing markets in Canada. An average detached house there now costs over C$2 million (USD $1.5 million). Data scientist Jens von Bergmann and UBC associate professor Nathan Lauster estimate that the metro area has 130,000 to 200,000 suppressed households: people doubling up with family or roommates who would live independently if they could afford to. That backlog alone represents six to ten years of recent construction. In 1981, two thirds of 25- to 29-year-olds in Vancouver lived in their own households; by 2021, only a third did. Rental vacancy rates often hover below one percent; vacancy rates in more affordable housing markets, such as Austin, Texas, hover around ten.

This pressure comes in part from Vancouver’s success. Vancouver is Canada’s third-largest metropolitan economy. With its stunning backdrop of mountains and ocean and its mild winters, it repeatedly ranks among the world’s most liveable cities. Accordingly, the metro area adds over 40,000 new residents each year, each needing somewhere to live.

Despite this massive demand for housing, and its dense urban core, most of the city looks as it did decades ago. Vancouver pioneered single-family zoning in Canada: the municipality of Point Grey became the first in the country to adopt such rules in 1922. When Point Grey amalgamated with Vancouver in 1929, the principle spread citywide, although initially in a relatively permissive fashion. In the 1960s, the city briefly experimented with allowing apartment towers in residential areas, but a political backlash voted in a council in 1972 that was more resistant to development. From that point onwards, Vancouver systematically downzoned, restricting building heights and locking most of its land into low-density use. Today, 52 percent of Vancouver’s residential land supports just 15 percent of its homes.

Vancouver never managed to comprehensively reverse the downzoning of the early 1970s, despite the huge pressure on housing affordability since. But it did find a way to permit intensive development in narrow strips, typically along transport corridors, while leaving the suburban expanses largely untouched. The city is dependent on such developments: property taxes are low, so almost half of its C$3.5 billion capital budget comes from developer payments. Through a mixture of fixed development fees and Community Amenity Contributions, which are negotiated during spot rezoning on a case-by-case basis, planners extract funding for parks, childcare centers, and affordable housing units in exchange for permission to build. They aim to capture 75–80 percent of the land value uplift. Because each negotiation process can be lengthy, the city prefers to allocate a small number of highly valuable permits, creating an incentive structure that favors concentrated towers, rather than gentle density across the city.

Subscribe to print

The downside is that the negotiation process is slow, expensive, and unpredictable. Von Bergmann notes that planners ‘typically get it wrong – they either charge too much or too little’. Getting it right requires predicting market conditions years in advance. Charge too much and the project is rendered unviable and dies, potentially taking the developer and many jobs down with it. Charge too little and the city misses out on valuable capital spending opportunities. Frances Bula, a journalist who has covered Vancouver planning for decades, describes the practical reality: ‘Big developers just cough up when planners demand.’ Smaller builders, including homeowners wishing to add small extensions to their own houses, lack the resources to play the game.

This system is what is known as Vancouver’s grand bargain, and what leads to the largely untouched suburbs occasionally punctuated with tall, thin towers. The grand bargain generates some income for public spending, while minimizing property taxes and development across most of the city. But the price is paid through worsened housing affordability and the economic costs of suppressed construction.

The mandate to build

After the land at Kits Point was returned to the Squamish Nation, they initially made plans to develop it at low density: two residential towers and some mid-rise offices, barely distinguishable in scale from the commercial buildings that already exist nearby. However, these plans stalled after the 2008 financial crisis, and it took some time for the Nation to return to its development ambitions.

In 2018, the Nation established a development arm, Nch’ḵay Development Corporation, bringing in much-needed expertise in construction. Moving quickly, in April 2019, they formally requested the Minister to order a referendum on the designation of the site, which Squamish people were eligible to vote in. In two votes, they approved the land designation for a mixed-use development, passing with 87 percent approval, and business terms with the development partner Westbank, one of Canada’s largest developers, passing with 81 percent approval. The Nation agreed to a 50-50 partnership with Westbank, though Westbank later sold its shares to OPTrust, which manages one of Canada’s largest pension funds. Despite a low turnout of 27 percent, the votes were considered to give a decisive mandate to the development plans.

But why did designating the land for development require a referendum? The site at Kits Point was returned to the Squamish Nation as reserve land. The Indian Act requires that reserve land be ‘designated’ by the federal government before it can be leased, a process that is considered ‘tedious, expensive and time consuming… with no guarantee of the outcome’, and which makes development difficult. Senakw’s swift progression is a testament to both political will and the Squamish Nation’s effective navigation of the system.

While reserve land can be leased with federal permission, it cannot be alienated from the band, meaning freehold ownership cannot be transferred in any way. This means it cannot be parceled up and sold off. It also means that the land cannot be borrowed against, because securing a loan on land is only possible if the land’s ownership can be alienated in the case of default. Traditional development finance uses loans as residential mortgages. Apartments cannot be sold as condominiums (where each unit is privately owned, like a house, and each owner also has a share of the overall building), but must instead use leasehold, where residents purchase a right to live in a property for a set period of time rather than owning their home outright.

Subscribe now

These restrictions might have made the project impossible without support from the federal government, which agreed to a C$1.4 billion loan through the Canadian Mortgage and Housing Corporation, the largest in the corporation’s history. To work within leasehold constraints, the residential units were earmarked for 100 percent rental.

Yet reserve land also offered a crucial advantage: because it is reserve land, it is exempt from zoning rules which apply to the rest of Vancouver. The Squamish Nation made the decision to ‘maximize economic benefits for the Squamish community’, which, in Vancouver’s housing-starved market, meant maximizing density: 6,000 rental homes plus commercial space. It also meant opting not to impose many of the costly mandates that cities typically require. They maintained the same safety and building standards that other Vancouver developments adhere to, but stripped away height limits, inclusionary zoning, zero-emissions requirements, parking minimums, design panels, and the years of public consultation that other projects must stick to.

Although Senakw is not subject to the normal city zoning rules, it nonetheless required standard public amenities like sewerage connection and transport access, including installing roads through public land. The City of Vancouver could have blocked or severely delayed the project simply by refusing these services, and might have been tempted to, given that the development sidesteps the city’s own planning rules.

Yet the city proved amenable. Kennedy Stewart, Vancouver’s mayor from 2018 to 2022, strongly supported the development and oversaw the signing of a services agreement in 2022. The Squamish Nation agreed to fund the sewerage upgrades, as well as the installation of cycle paths and seawall repairs to reduce flood risk to the neighborhood, paying roughly what they would have for site infrastructure under the standard planning process. They were not, however, required to pay the far larger sums that Vancouver normally extracts from developers to fund other municipal priorities.

Mayor Stewart emphasized to me that Canada’s intensifying political discourse around ‘reconciliation’ was critical. Many dismissed this discourse as performative symbolism, but at Senakw it had practical consequences: ‘Promises about reconciliation forced governments’ hands when it came to enabling indigenous-led developments’. The Squamish Nation’s referendums gave the project a democratic legitimacy that Vancouver’s usual stakeholder consultations rarely achieve. A city government that had spent years resisting modest density increases facilitated a far more ambitious project when backed by a clear community mandate and a compelling moral case.

The financial rewards to the Squamish Nation are substantial. A minority of the homes are earmarked as below market rate housing that will be let out to members of the Squamish Nation. The rest will be let out at market rates. With their 50 percent stake expected to generate up to C$10 billion in rental income over the entire lifespan of the buildings, and around 4,000 members, the project could be worth around C$2.5 million per person. How it will be distributed or spent is as yet undecided.

Beyond Senakw

On just one site, Senakw will provide a year’s worth of housing normally built in the city of Vancouver. It functions as a sort of policy laboratory, showing that cities like Vancouver can build many more homes if the decision-makers allow it.

Senakw also offers two lessons for those of us in other cities who want to get more homes built. The first is about incentives. When the Squamish Nation’s members voted on development, they were voting on their own land, and for the economic development of their own community. They chose to maximize density and limit the costly requirements that cities like Vancouver typically impose on new developments because they would bear the costs of these requirements directly. Yet in terms of quality and safety, Senakw is just as good or better than other new developments built just across the river. But residents voting on their own projects make very different assessments of the value of some of these mandates and vote accordingly. Similar dynamics have appeared in other resident-led ballots such as estate regeneration programs in England, and Israel’s Pinui Binui program.

Share

The second lesson concerns democratic legitimacy. It is partly a myth that Senakw slipped through a loophole in normal land-use restrictions. The City of Vancouver could have obstructed Senakw by refusing infrastructure connections and services. The federal government could have obstructed it by refusing the land designation or not offering the financing agreement. That both chose not to do so reflects the power of a clear democratic mandate. Governments that spend years resisting modest density increases may nonetheless facilitate ambitious projects when backed by direct community approval and a compelling moral case.

It’s an open question how replicable Senakw’s model is in the rest of Canada. Most reserve land in Canada is not urban; the Indian Act ensured that most such sites were seized long ago. However, ongoing legal battles make the future unpredictable. Indigenous groups may apply to the federal government to convert land they own to reserve status, and at the moment the federal government is amenable to doing so.

Nations may also build on land they own outside of reserves using standard zoning processes. Not far away from Senakw sit the Jericho Lands, a 90-acre site being developed in a joint partnership of the federal government and the Musqueam, Squamish and Tsleil-Waututh Nations. This site is owned on a freehold basis. The overall development plan has been approved through the standard city planning process, although it will require further approvals of rezonings on individual parcels. It is expected to take 25–30 years to deliver around 13,000 new homes, an annual rate much slower than Senakw.

Subscribe to print

Whatever lessons cities do or do not draw from Senakw, for the Squamish Nation, the significance extends beyond housing policy. For Wilson Williams, it represents something important: ‘We’re no longer out of sight and out of mind in our own village.’ More than a century after his ancestors were forced to sail away from the burning ruins of their homes, a new generation of Squamish people will return to live where their ancestors once fished for salmon in the abundance of False Creek.

Anya Martin is policy lead at YIMBY Alliance.

Artwork by Nina Bunjevac.

The author would like to thank for their generously given time and expertise: Squamish Nation, Abundant Housing Vancouver, Revery Architecture, Jens von Bergmann, Randy Bouchard, Frances Bula, Dorothy Kennedy, Nathan Lauster, Vanna Orecchio, Kennedy Stewart, Sam Sullivan and Lynn Zanatta.

Cities are often centers of agglomeration, places where people gather to collaborate with one another. But this is not the only reason they exist. Sometimes, cities are chiefly centers of consumption, where elites gather to devour resources extracted from the rest of the country. And occasionally, they are something like prisons, where troublesome social groups are concentrated so that the authorities can keep an eye on them. Many premodern cities, like Versailles, Naples, or Imperial Rome, were a little like this. But perhaps the greatest example was Tokugawa Edo.

Between 1600 and 1868, Japan was dominated by the Tokugawa family. The Tokugawas had prevailed over their rivals after a series of civil wars, establishing a sort of dictatorship known as the Shogunate. They developed a remarkable social system, crafted to preserve their power, and with it, the peace and social stability of Japan. At the apex of this system was the city of Edo (today’s Tokyo), at times the largest city in the world, and one of the strangest urban structures in history.

Subscribe now

The Tokugawa social system

In early modern Europe, most people were tenant farmers, who paid rent to landowners. The state sometimes taxed landowners, sometimes tenants, and sometimes both through consumption taxes. In peacetime, however, the early modern state did not do very much, so taxes generally ran at just a few percent of national income.

The picture in Japan was profoundly different. The peasantry was directly taxed by the government, at rates varying from 15 to 70 percent of the harvest, with 40 percent as a rough norm. The authorities distributed most of their tax receipts to the samurai, a hereditary, quasi-noble class making up about six percent of the population. In both cases, the agricultural surplus ended up in the hands of a leisure class, but the Japanese system was structured very differently, with the surplus reaching the leisure class only through the funnel of public taxation.

Measured by agricultural output, about 15 percent of Japan fell under the direct control of the Shogunate. These were areas that had always belonged to the Tokugawa, or that had been conquered by them during the civil wars. About a tenth of samurai were Tokugawa retainers, who received their stipends directly from the Shogunate.

In about three quarters of Japan, the tax authorities were a different group, the daimyo. The daimyo had originally been the territorial nobility, and in the pre-Tokugawa era, the breakdown of central authority had left them as the effective rulers of their domains. After 1600 the surviving daimyo submitted to the Tokugawa and were rewarded with a role somewhat akin to that of regional governors. There were about 260 daimyo, and about nine tenths of samurai were their retainers.

Urban design as a panopticon for the nobility

The city of Edo played a crucial role in this system. The daimyo were required to maintain mansions in Edo, in which their families were obliged to live permanently. Any act of disloyalty by a daimyo would thus place his family in the gravest peril. Most women of the daimyo class passed their lives as effective hostages of the state, never visiting the domains that their husbands, fathers, and sons governed. The daimyo themselves were also required to spend alternate years or half-years in Edo, bringing with them great crowds of samurai retainers. Edo thus had a dual nature: on the one hand, it was the apex of Japanese society, in which the country’s agricultural surplus was consumed; on the other, it was a kind of prison, in which Japan’s potentially dangerous elites were contained and monitored.

This gave Edo a peculiar demographic and economic character. For one thing, it was enormous. Japanese rice agriculture was extremely productive by premodern standards and generated a huge surplus for the capital’s benefit. Edo’s population seems to have been over a million by 1700, which would make it the largest city in the world at the time: London reached a population of a million only around 1800, while New York did not reach it until 1880.

Subscribe now

It was also extremely top-heavy socially. Unlike the knights of the European feudal system, Tokugawa-era samurai were usually required to live in cities. The Shogun’s own samurai were permanently concentrated in Edo, while each daimyo brought hundreds or thousands of his samurai with him during his years of residence in the capital. This meant that almost half of Edo’s population at any given time were samurai.

The samurai were theoretically a warrior class, but since Tokugawa Japan was at peace, there was little real soldiering to do. Most non-military work was seen as degrading, with the exception of gentlemanly occupations like civil administration and tutoring in calligraphy. Most samurai thus lived as pensioners of the state, deriving the bulk of their income from hereditary government stipends assigned to their families at the start of the Tokugawa era.

Stipends were generous for high-ranking samurai but very modest for the majority: in 1876 the Japanese state calculated that only 5.1 percent of samurai had stipends worth more than 100 yen, equivalent to a little over $4,000 today. The result was that most samurai lived in dignified but extreme poverty. It is a testament to the primacy of status among human ends that, although some samurai surreptitiously took on side jobs to supplement their tiny incomes, very few abjured their rank to take up more gainful occupations openly. Interclass marriages were forbidden, and there was theoretically no process for ennoblement, though some wealthy commoners worked around this by arranging to be adopted into impecunious samurai families.

Subscribe to Print

The High City

Edo was zoned according to social class, probably the largest-scale use of zoning before modern times. Samurai neighborhoods made up a large minority of the total surface area, concentrated in western areas that came to be known as the High City. Higher-ranking samurai had substantial homes, but most lived in extremely austere conditions: the median family lived in a tiny, two-room rowhouse with no garden and communal latrines.

The samurai shared the High City with the daimyo. Each daimyo was expected to maintain several estates in Edo, including a more compact estate near the Shogun’s castle and a more spacious garden estate further out. These were made up of single-storey buildings connected by covered walkways, set amid exquisitely landscaped gardens. Each estate was surrounded by a perimeter wall broken by just one or two richly ornamented gates. The great expense of all this was, from the Shogunate’s perspective, part of its appeal: the more resources daimyo expended on maintaining mansions and circulating between them, the less remained for fomenting trouble.

These samurai and daimyo neighborhoods made up a substantial majority of Edo’s surface area. In some ways, they foreshadowed affluent American suburbs, though they would have felt quite different at street level. Physically, the closest modern parallel might be elite neighborhoods in countries like South Africa or Brazil, where fear of crime has driven the wealthy behind high walls and electric fences.

In fact, however, Edo seems to have been an unusually safe city, a fact that often struck early Western visitors. Most people had no locks on their front doors. The English writer Isabella Bird wrote that a solitary foreign woman could travel in Japan in ‘perfect safety’ and that she had never encountered ‘a single instance of incivility’, though at home she would often have been ‘exposed to rudeness, insult, and extortion, if not to actual danger’. The walls and gates of the High City may thus tell us more about elite culture and status display than they do about any distinctive security needs.

Share

The Low City

Much of the rest of Edo’s surface area was taken up with religious complexes and institutional buildings belonging to the Shogunate. The commoners were crowded into the remainder, mostly in riverside areas known as the Low City. Though most commoners were very poor, some were richer than most samurai, and even richer than minor daimyo. Some of the greatest artistic legacies of Tokugawa Japan, like ukiyo-e prints and the kabuki theater, arose from the patronage of this class. However wealthy they might become, however, commoners’ status was axiomatically lower than that of the samurai.

Edo’s commoners worked mostly in providing goods and services to the city’s elite. The city as a whole produced little, consuming resources from the rest of Japan and providing little in return save the arguable gift of government. No international trade took place in Edo: the government was intensely wary of destabilizing foreign influences, and by the late-eighteenth century it allowed only two Dutch and ten Chinese ships to visit Japan each year, all through the carefully controlled port at Nagasaki. (For comparison, about fifteen thousand ships arrived in British harbors from foreign ports annually in the same period.) In this as in other respects, the Shogunate’s political needs had a decidedly fitful relationship to the country’s economic growth: although the political stability that the regime provided certainly did foster prosperity, the steps it took to ensure that stability often stifled it.

The physical form of the Low City was even more remarkable than that of the daimyo neighborhoods. Commoner Edo was divided into some 1,500-2,000 fenced and gated blocks called chō. These were then subdivided into roji, alleys lined with small houses, often also gated. The Low City was thus divided up by many thousands of internal checkpoints, all of which closed at night. Edo was not under threat of attack in the Tokugawa period and the city as a whole was not fortified. The purpose of this immense labyrinth of walls and gates was to control and monitor the movement of the population.

Wealthy merchants lived in substantial houses called machiya, featuring a series of small garden courtyards around which living spaces were ingeniously laid out. In plan, machiya recall the domus house type of the urban middle and upper classes in the Roman Empire. Many machiya must have been rather agreeable houses, and some Japanese people still live in modernized machiya today. Most commoners, however, lived in nagaya, tiny terraced houses frequently built back-to-back. A standard specification for nagaya was 2.7 meters wide and 3.6 meters deep, yielding a total floorspace of only 13.2 square meters.

An enormous number of people were crammed into commoner neighborhoods. The most crowded had 58,000 inhabitants per square kilometer, about twice the population density of Manhattan today. Remarkably enough, however, their dwellings were largely single-storey. This seems to have been required by regulation, in turn motivated by fire risk or by dislike of commoner buildings towering over the mansions of the nobility. Japanese builders were perfectly capable of erecting taller structures, so the fact that these intensely crowded neighborhoods forewent the floorspace that additional storeys could have provided is not least among the paradoxes that Tokugawa Edo presents to history.

Subscribe now

* * * *

In a way, the Tokugawa system was a success. Japan experienced near-total peace between 1600 and the late nineteenth century, a remarkable achievement for a premodern society and a dramatic contrast to Europe or China, where tens of millions of people died in wars. The Tokugawa succeeded in bribing, blackmailing, and intimidating Japan’s armed elites into a quarter of a millennium of passivity, enabling steady economic growth and a remarkable artistic flowering. One might thus argue that Edo’s apparent parasitism was an illusion. Edo was indeed a gilded prison, but prisons can be useful things.

Still, it would have been nice if Japan had not needed a vast prison capital in the first place. Tokugawa Edo stands as a monument to the power of rent-seekers, producing little and demanding immense resources as a condition of civil peace. It shows how the physical form of cities may be reshaped by these demands, as governments apportion space and limit movement in line with their political needs. These forces yielded a city with tight restrictions on built density and land use, an acute housing shortage, and a huge share of privileged state stipendiaries. Edo is a particularly striking example of these phenomena. But it is far from alone.

Samuel Hughes is an editor at Works in Progress. Follow him on Twitter.

Thanks to Fabian Drixler for generous help.

In the twentieth century, Christian and Jewish religious architecture was reshaped by modernism. Most new churches and synagogues today are built in modern styles; almost no new Gothic cathedrals have been built since the nineteenth century. Hindu architecture did not undergo the same process. Temple architecture is governed by ancient canons known as the Shilpa Shastras, which define the classical style of Hindu temple architecture. In the twentieth century, many Hindu communities continued to aspire to this style, while sometimes using modern materials for reasons of cost. As these communities have become more prosperous, they have tended to adhere to the these Shastras even more closely: modern affluence has enabled a more faithful embodiment of architectural tradition rather than a departure from it. Today, these monumental projects form the world’s foremost living tradition of masonry architecture, united to vast global supply chains and great feats of logistics and engineering.

To understand the Hindu temple, one must invert the modern conception of a public building. A temple is not an auditorium designed primarily for the convenience of a human audience: it is a residence of God, who is manifested on Earth through a consecrated image (mūrti) accommodated in the temple. Within Hindu theology, the sacred image is not a statue, but God’s living presence, which the architecture of the temple is designed to revere. Its scale, masonry, and ornamentation collectively aim at creating a residence befitting the majesty of God, while generating the peace and tranquility associated with the divine.

Temple architecture normally follows rules set out in the Shastras, architectural treatises composed between the fifth and fifteenth centuries. The Shastras function both as artistic guidelines and rigorous algorithms. The measurements of the temple’s components, from the height of the plinth to the curve of the shikara (the tower or spire), are calculated to correspond to the geometries of certain constellations.

The Shastras also rigorously govern construction methods. The basic structure of most large modern buildings is a steel or reinforced concrete frame: if stone is present at all, it is merely a surface cladding. In Hindu temples, by contrast, the structure is still formed of massive load-bearing stones. While steel is sometimes introduced in foundations and structural cores to meet modern seismic standards, the governing structural logic remains resolutely premodern. Stability is achieved through the compressive strength of stone and through complex interlocking jointing systems, including mortise-and-tenon-like connections.

This approach prioritizes longevity over speed. Steel frames are strong, but they often succumb to corrosion over time. This is one reason why many office buildings have a design working life of 50 years, after which they are sometimes replaced. Load-bearing masonry is slow to build, but it is extraordinarily durable. Temples built today could stand for thousands of years, vanishing only when their stones are worn away by the wind and the rain.

Perhaps the most striking innovation in recent temple architecture is the labor model. The construction of a modern temple can be a complex hybrid of high-tech manufacturing, hereditary craftsmanship, and devotional volunteerism. Stone is sourced from quarries around the world, especially Tuscany, Bulgaria and Rajasthan. It is shipped to workshops in India, where intricate carving is executed by professionals. These artisans often come from lineages of stone carvers who have preserved these skills for centuries.

The carved stones are then shipped to the construction site, a massive logistical puzzle involving tens of thousands of numbered pieces. Their assembly operates on a unique collaborative model: the more technical work is generally executed by professionals, but other tasks are performed by a host of volunteers. The construction process itself is often viewed as a form of worship: tools are sanctified and prayers mark critical stages, ensuring the sacred character of the temple is maintained throughout its creation. This integration of hereditary skill with devotional energy creates a site that can feel less like a construction zone and more like a festival.

Subscribe now

Here are seven examples of this living tradition. They are particularly impressive buildings, but they are not fundamentally exceptional: smaller and simpler temples in the same style are being built today in thousands of towns and villages across India.

1. Shri Swaminarayan Akshardham

Gandhinagar, Gujarat, India

Construction: 1979–1992

The BAPS Swaminarayan Akshardham in Gandhinagar is a pivotal building in modern Hindu architecture. Constructed over a period of 13 years, it is built from 6,000 tons of pink sandstone from Rajasthan, chosen for its durability and aesthetic warmth. Standing 108 feet tall, 240 feet long, and 131 feet wide, the structure is a complex assembly of 97 intricately carved pillars, 17 domes, 220 stone beams, and 256 statues representing the breadth of the Hindu spiritual pantheon.

The interior layout is designed to facilitate the dissemination of knowledge. The upper floor gallery houses a comprehensive series of paintings illustrating the biography of Bhagwan Swaminarayan, a visual text that makes his life and philosophy accessible. The lower floor serves as a repository for historical relics – personal objects used by Swaminarayan and preserved by his followers.

2. BAPS Hindu Mandir

Abu Dhabi, United Arab Emirates

Construction: 2019–2024

The BAPS Hindu Mandir in Abu Dhabi is the first traditional Hindu stone temple in the Middle East (‘mandir’ means ‘temple’ in Hindi). Built on land gifted by Sheikh Mohamed bin Zayed Al Nahyan, the temple was constructed using pink sandstone from Rajasthan for the exterior and Italian marble for the interiors. The temple features seven towering shikharas (spires), housing seven revered deities from across India, as well as representing the seven emirates of the UAE. Upon entering, visitors encounter streams representing the holy rivers Ganga and Yamuna, while a beam of light symbolizes the mythological Saraswati River.

Construction in the desert environment presented specific engineering challenges. The temple is the first traditional Hindu temple to undergo complete digital modeling and seismic simulation to ensure resilience against local geological conditions. The foundation utilizes a concrete pour reinforced with fly ash to reduce its carbon footprint.

Inside, 402 uniquely carved marble pillars support the structure, which includes a ‘Dome of Harmony’ depicting the five elements of nature and a central ‘Dome of Peace’. A notable feature is the stone-carving symbolism. While Hindu narratives adorn the exterior, to reinforce the message of Vasudhaiva Kutumbakam (‘the world is one family’), the carvings also depict moral stories from Arabian, Mayan, Greek, and other ancient civilizations.

3. BAPS Shri Swaminarayan Mandir

London, United Kingdom

Construction: 1992–1995

Popularly known in Britain as the ‘Neasden Temple’, the BAPS Shri Swaminarayan Mandir was the first traditional Hindu stone temple in Europe. Constructed in less than three years, the temple required five thousand tonnes of Italian Carrara marble and Bulgarian limestone. The limestone was specifically selected for its density and durability to withstand London’s damp climate and urban pollution. The stone was shipped to India for carving, after which over 26,300 pieces were shipped to London. The specific architectural layout corresponds to the Revati Nakshatra, a lunar constellation in Hindu astrology, ensuring the design aligns with precise spatial and auspicious parameters.

The foundation itself marked a complex engineering event. At the time, it was one of the largest single concrete pours in the UK, creating a massive raft capable of supporting the immense load of the stone superstructure. The project was financed entirely by the local community through grassroots efforts, including sponsored walks and aluminum can recycling campaigns. Over 3,000 local volunteers worked on the construction itself. Beyond its function as a place of worship, the complex regularly hosts school visits and features exhibitions on Hinduism and community outreach programs. Visitors have included royalty and prime ministers.

4. Swaminarayan Akshardham

New Delhi, India

Construction: 2000–2005

Dedicated to Bhagwan Swaminarayan, BAPS Swaminarayan Akshardham is a significant example of contemporary traditional stonework. Rising 141 feet, spanning 356 feet in length and 316 feet in width, the temple was completed in five years through a collaboration of 11,000 artisans and 8,000 volunteers. More than 300 million volunteer hours went into making the complex. The temple was originally envisioned by Yogiji Maharaj (1892–1971) as a temple on the banks of the sacred River Yamuna, and completed under the leadership of his successor, Pramukh Swami Maharaj (1921–2016).

The structure rests upon the Gajendra Peeth, a three thousand ton plinth featuring 148 sculpted elephants. These depict storied elephants from various texts and pay homage to the foundational support of the animal kingdom.

Subscribe now

Above them rises the temple’s intricately carved outer wall, the largest such wall built in India in eight centuries. It depicts animals like lions and elephants, followed by floral motifs, culminating in a layer adorned with various sculptures of deities and sages. The temple itself incorporates 20,000 statues across its 234 pillars and nine ornate domes. The pillars and ceilings depict various incidents and virtues from the life of Bhagwan Swaminarayan. The complex extends beyond the sanctuary to include educational facilities, featuring exhibitions, a cultural boat ride, and a multimedia water show depicting events from the Upanishads, some of the central Hindu religious texts.

5. Pramukh Swami Maharaj Smruti Mandir

Sarangpur, Gujarat, India

Construction: 2018–2023

The Pramukh Swami Maharaj Smruti Mandir represents a specific category of Hindu temple architecture known as a Smruti Mandir, or memorial shrine. Located in Sarangpur, this temple marks the cremation site of Pramukh Swami Maharaj (1921–2016), a significant figure in modern Hinduism who oversaw the construction of more than 1,200 temples, including the Akshardham complexes. Within Hindu traditions, sites associated with the final rites of enlightened spiritual leaders and virtuous figures are sacred spaces.

The temple is constructed entirely from pristine white Makrana marble, the celebrated Rajasthan stone used in the Taj Mahal. It features a single central dome and detailed carvings that illustrate the narrative of the guru’s life and global travels. The temple functions as a canopy over the specific site of the funeral rites. The original cremation spot has been preserved in its raw state within a transparent casing alongside the ashes, allowing visitors to view the ground directly. The current spiritual head, Mahant Swami Maharaj, inaugurated the temple with the stated objectives of fostering peace, faith, and the fulfillment of personal wishes.

6. BAPS Shri Swaminarayan Mandir

Chino Hills, California, United States

Construction: 2004–2012

The BAPS Shri Swaminarayan Mandir in Chino Hills presents a specific engineering application: reconciling the rigidity of traditional stone architecture with the safety requirements of a high-risk seismic zone. Located in the San Bernardino region, an area known for frequent tectonic activity, this structure is the first traditional Hindu temple designed with a base-isolation system. The exterior features pink sandstone from Rajasthan; the interior uses white Italian marble from Carrara.

To protect this brittle stone assembly from ground motion, the structure rests on a system of 40 base isolators – consisting of steel plates stacked with a viscous liquid – which mechanically separate the upper temple from its foundation. This mechanism is designed to allow the heavy edifice to shift laterally during an earthquake, effectively enabling it to ‘float’ independently of the ground. Engineers and architects refer to this configuration as the ‘floating mandir’.

The project required navigating complex regulatory and height restrictions. Temple representatives engaged with local officials and residents to explain the cultural and religious significance of the traditional spires to secure approval. Beyond the temple itself, the complex includes a cultural center, which houses a museum dedicated to the history and details of Hindu architecture. The site operates largely on solar energy, merging traditional aesthetics with modern ecological standards.

Subscribe now

7. Birla Mandir

Kolkata, West Bengal, India

Construction: 1970–1996

The Birla Mandir in Kolkata was sponsored by the industrialist Birla family, whose name it bears. The construction took 26 years. While the temple is consecrated to the Hindu deities Lord Krishna and his consort Radha, the complex also houses shrines for other major deities including Shiva, Durga, and Shakti.

Constructed from white marble and cream-colored sandstone, the edifice reflects a synthesis of regional aesthetics. It draws primary inspiration from the Lingaraj Temple in Bhubaneswar while incorporating elements of Rajasthani architecture and modern styling. The design principles are rooted in the Brihat Samhita, an ancient Sanskrit text that outlines precise standards for temple architecture and iconography. Standing 160 feet tall on a raised platform – a classical feature intended to separate the sacred space from the street – the structure is a defining landmark of the Kolkata cityscape.

The temple’s interior and exterior are notable for their didactic ornamentation. Intricate carvings depict pictorial narratives from Hindu philosophy and scenes from scriptures like the Bhagavad Gita. The project sustained a hired workforce of artisans from the city of Muzaffarpur in Bihar for over two decades, playing a crucial role in preserving traditional masonry skills.

Tilak Parekh is a researcher of religion and anthropology at the University of Cambridge.

The phones we carry around in our pockets have two million times more memory and are thousands of times faster than the room-sized computers that guided the Apollo mission to the Moon. This incredible shrinking act has been driven by our ability to make transistors smaller and smaller.

Each transistor is a microscopic switch that can alternate between a one and a zero, the basic language of all computing. Billions are packed onto tiny silicon chips called semiconductors. The more transistors that fit onto a chip, the more logic and memory circuits it holds, and the more it can do.

Advanced semiconductors are, arguably, the most important technology in the world. Over the last five years, they have even emerged as a geopolitical flashpoint between the US and China. But for all this rivalry, any country or company that hopes to manufacture semiconductors is dependent on a single firm: ASML. Dubbed ‘a relatively obscure Dutch company’ by the BBC in 2020, ASML makes the only machines in the world capable of stenciling the transistors onto chips with the precision necessary to fit billions on a 30-centimeter wafer.

These machines are roughly the size of double-decker buses. To ship one requires 40 freight containers, three cargo planes, and 20 trucks. They are the world’s most complex objects. Each contains over one hundred thousand components, all of which have to be perfectly calibrated for the machine to produce light consistently at the right wavelength.

While ASML is now the sole supplier of these machines, and will be for some time to come, it started out as a laggard in the chipmaking industry. Overtaking its competition required many things rarely associated with European companies: close collaboration with the American government, selling large stakes to foreign competitors, and a huge gamble on an unproven technology.

Let there be light

The key to ASML’s success is a technology called photolithography (sometimes just called lithography). The technique involves transferring a pattern onto a semiconductor wafer by exposing it to light. In the 1950s, the first chipmakers had tried to draw these patterns by hand, but anything that physically touches the wafer scratches it, dirties it, or warps the pattern. Scientists working independently for Bell Labs and the US military realized that they could use light to print identical patterns without making physical contact with the wafer.

To make chips, engineers start with a thin wafer of semiconductor material, usually silicon. This wafer is coated with a chemical called photoresist, which reacts when exposed to light. In photolithography, light is projected through a detailed pattern onto the photoresist-coated wafer, softening the exposed areas. The wafer is washed to remove any softened areas, revealing the silicon underneath. It is then moved to an etching machine that blasts it with charged chlorine or bromine gas, carving the desired pattern into the exposed silicon. These features are later filled with metal, such as tungsten and copper, to connect the transistor to power. These etched layers then combine into an intricate network of transistors.

Over time, the semiconductor manufacturing ecosystem has developed increasingly sophisticated etching using ever smaller wavelengths of light. Smaller wavelengths diffract less, allowing the light to travel in straighter lines and print sharper, tinier details without blurring. These allow for more precise pattern projections that, in turn, allow smaller and more densely packed transistors.

Early lithography relied on mercury vapor lamps that were similar to streetlights, while more modern machines rely on lasers created using argon and fluorine gases. By 2010, such lasers made it possible to create a 22-nanometer feature through multiple exposures using a 193-nanometer wavelength.

The most advanced version of this technology, extreme ultraviolet lithography, is used to make the very smallest chips. The smallest in 2025 were marketed as three nanometers, roughly 25,000 times thinner than a human hair.

To make them, a droplet of liquid tin is released into a chamber and hit with a single pulse of light, which melts and flattens it. As the droplet continues to fall, a second, more powerful pulse vaporizes the tin, creating an extremely hot plasma that emits light at the narrow wavelengths needed for extreme ultraviolet lithography. The light beam is then concentrated by reflecting it across a series of slightly concave mirrors so flawless that, if scaled to the size of Germany, their imperfections would be measured in millimeters. Engineers need to use mirrors, rather than the glass lenses used in standard lithography, as almost all solid materials absorb light at such short wavelengths.

The light eventually hits the mask, which contains the pattern to be printed on the chip. As the pattern on the mask is usually several times larger than what is wanted on the chip, the light is then reflected by a second system of mirrors.

After the light reflects from the mask, it carries the pattern as a bundle of rays spreading out from each point. The next mirrors tip these rays inward so that, instead of spreading widely, they reunite over a shorter distance. When the rays from each point come together sooner, the picture they form is physically smaller. By repeating this with several carefully shaped mirrors, engineers shrink the pattern by a fixed amount while keeping it in focus. After being shrunk four times, it hits the wafer.

Share

The great shrinking act

Longer wavelengths act like a blunt chisel, suitable for rough shaping, but they struggle to capture finer details. The longer light waves are larger relative to the tiny features on the reticle that they must reflect from. When a wave meets something smaller than itself, it naturally spreads and bends around its edges instead of casting a sharp shadow. To create the same details, the blunt chisel needs to go over the same spot a number of times (creating blurrier edges). Lithography had to take wavelengths all the way to the extreme ultraviolet range to achieve the high resolution patterning needed for cutting-edge process nodes.

Wavelengths as low as 13.5 nanometers can achieve more precise patterns in a single exposure. In fact, extreme ultraviolet lithography can combine three or four photolithography patterning cycles into a single one on a seven-nanometer node. Without EUV, producing five-nanometer nodes might require as many as one hundred different steps.

Extreme ultraviolet lithography was able to produce more accurate patterns on wafers than older techniques even if they were used multiple times.

Today, ASML dominates the overall market for lithography and has an effective monopoly in extreme ultraviolet lithography. Its EUV machines sell for more than $120 million. With a market capitalization of over $400 billion, ASML is one of Europe’s most valuable companies. But it wasn’t always like this.

Subscribe now

Origins

ASML started off life within Philips, the Dutch consumer electronics giant. During the 1970s, Philips had roughly 20 percent of the global electronics market and was a major chipmaker. In this era, lithography machines used wavelengths of over 400 nanometers to pattern 1,000-nanometer features. The industry struggled to shrink features without losing accuracy or letting dust and flaws creep in. Philips began to work on its own prototype, drawing on its expertise in optics and precision mechanics. By the early 1980s, the project was running into trouble. The company was looking to cut costs and engineers estimated that they would need over $280 million in today’s money to finish the machine’s development and production.

In 1984, Philips spun out Advanced Semiconductor Materials Lithography (which later dropped the full name in favor of its acronym) as a joint venture with ASM International, a Dutch conglomerate that sold equipment to the semiconductor industry. The business originally struggled. It had no market share and no brand recognition. Its first product, the PAS 2000, was a commercial failure. The machine used oil pressure, like that in power steering, to move the table that held the wafer during exposure, rather than electric motors. This made it smooth and precise, but it was prone to leaking. At the first conference ASML attended, one industry executive told them: ‘The race has already been run. There’s no room for you here.’ ASML switched back to electric motors.

The company took an unusual approach from the outset. While Japanese giants Nikon and Canon were vertically integrated, ASML outsourced key components like optics and motors so that it could focus on assembling and optimizing the final machine. Given this outsourcing, it made sense for ASML to embrace a modular design with clearly defined subsystems. This approach was mocked in European manufacturing circles. German engineers warned ASML’s leadership that they were ‘asking for trouble’ and would ‘lose all control’ if they didn’t make critical components themselves. But ASML had no choice: it lacked the capital, expertise, and time to build these subsystems from scratch.

By 1988, ASML was on the verge of collapse. ASM International had already pulled out, and Philips considered shutting it down. It was saved by a single Philips board member, Gerd Lorenz, who was particularly worried about Europe’s growing dependence on Asia for strategic technology. Lorenz argued that Europe needed a stake in chip manufacturing. This was enough to convince Philips to give ASML more time, but didn’t fix its fundamental problem: it was still an inferior supplier with no competitive edge.

ASML used the time it was given to develop the PAS 5500, released in 1991 and the company’s first commercial breakout. While Nikon’s contemporary photolithography system was more precise, ASML’s modular design meant that machines could be fixed quickly on site. This reduced downtime and, by making it easy to replace parts when they broke, it was possible to extend the machine’s life. This was a key factor that led John Kelly, IBM’s director of semiconductor R&D, to push IBM to order the PAS 5500 over the Japanese machines. ASML had gone global.

The first breakthroughs

ASML’s success depended on two projects in the late 1990s and 2000s that gave it a huge advantage in research and development. The first was a public-private partnership, started in 1997, called the Extreme Ultraviolet Limited Liability Company. The Extreme Ultraviolet Limited Liability Company began life as a rescue mission. Before 1997, basic semiconductor research was carried out in a small handful of research labs, all dependent on government grants.

The original program for EUV research was a ‘virtual national lab’ that combined Lawrence Livermore National Laboratory, Sandia National Laboratories, and the Lawrence Berkeley National Laboratory. Each covered a different component: Livermore focused on mirrors and optics, Sandia on the light source and systems engineering, and Berkeley on advanced equipment for testing. But in 1996, Department of Energy budget cuts had placed the virtual national lab program on the chopping block.

Intel, then the undisputed world leader in microprocessors, was keen to preserve the work and spearheaded the creation of the Extreme Ultraviolet Limited Liability Company, the largest public-private partnership of its kind in the history of the US Department of Energy. During its six-year life, the company invested over $270 million into extreme ultraviolet lithography development, funded by the sale of shares to member companies, giving them a right of first refusal to purchase the photolithography tools being produced.

The company initially restricted membership to American firms. ASML, along with its main Japanese rivals, Canon and Nikon, was initially barred from membership.

The only established semiconductor equipment manufacturer to join the partnership from the beginning was Silicon Valley Group, which had a market share of just 5 percent to ASML’s 20 percent. Fearing the danger of being reliant on such a small manufacturer, the rest of the companies involved concluded that it would be better to open up to foreign firms, rather than risk ceding the entire market.

ASML was allowed to participate so long as it committed to establish a research center in the US and source 55 percent of components for the systems sold in the US from American suppliers. In practice, this commitment was never enforced. Its Japanese competitors were never allowed to join, due to widespread fear in the US of Japanese competition.

The program built up a vast base of intellectual property and process knowledge. These types of public-private partnerships typically grant the participating companies a non-exclusive license to use the intellectual property generated, but in this case the companies in partnership got complete ownership.

In 2001, ASML acquired Silicon Valley Group after it ran into cash flow difficulties, making ASML the sole surviving equipment manufacturer in the partnership. When the consortium produced the first full-scale extreme ultraviolet lithography prototype – the Engineering Test Stand – ASML stood alone at the vanguard of lithography. This was the first demonstration that 13.5-nanometer light could print dense features on a chip.

By the time the Engineering Test Stand was built, the program had already proved that it was possible to generate extreme ultraviolet light reliably, which let engineers start building mirrors and lenses that could be used in real production tools. To solve outstanding questions, such as how to boost the throughput of their machines or increase the power of their light sources in production settings, ASML needed to test its machines in environments close to the real world. But no chipmakers were willing to shoulder a project so large and risky at such an early stage.

The second project essential to ASML’s success was the Belgium-based Interuniversity Microelectronics Centre (IMEC), a research organization that collects machines from different companies and allows researchers to test them in semi-real environments while protecting the companies’ intellectual property.

As potential customers began to consider different options for next generation lithography technologies, ASML used IMEC to promote its extreme ultraviolet lithography prototype. Topping ASML’s target list was TSMC, which today is the world’s largest semiconductor foundry. Founded in 1987, TSMC’s history had been intertwined with ASML’s since its birth: Philips, ASML’s former parent, owned a 27.5 percent stake in it. Seeing ASML’s machinery exhibited at IMEC was what led TSMC to partner with ASML in EUV development.

By contrast, Canon and Nikon were tight-lipped about their research and made little effort to cooperate with outside companies. While this theoretically allowed them to maintain greater control over their work, and capture more of the value chain, it also made them solely responsible for simultaneously solving a bewildering array of fundamental physics problems, while assuming all the financial risk of doing so.

Since almost all of the parts in ASML’s machines are made by other companies, it has become master of a sprawling supply chain of over five thousand companies. It has diversified its suppliers over the years in a very deliberate way: 80 percent of its spending goes to companies across Europe and the Middle East (notably not the US, despite prior agreements), which reduces the risk of potential export restrictions, tariffs, and other geopolitical risks that may face critical suppliers based in the US or Asia. It also aims for its suppliers to make no more than 25 percent of their revenue from ASML, to force them not to become overreliant on the volatile semiconductor market.

While most of its components come from a large number of small suppliers, ASML has formed deep bonds with its biggest suppliers. It acquired a 24.9 percent stake in optics manufacturer Zeiss. Peter Leibinger, vice chairman of laser manufacturer Trumpf, has said that ASML and Trumpf are a ‘virtually merged company’.

Winning the war

Extreme ultraviolet lithography would not become a successful commercial technology until 2018, over 20 years after the creation of the Extreme Ultraviolet Limited Liability Company and 34 years after IMEC was founded. In the meantime, it was consuming more and more resources. By 2015, ASML was spending more than $1 billion a year on R&D, more than double its 2010 total. According to some estimates, by 2014, the industry had collectively invested over $20 billion in extreme ultraviolet lithography, with no guarantee of any return.

ASML was able to continue pouring money into this black hole partly because it had already beaten its competitors. By 2010, it had two thirds of the overall lithography market and was the dominant supplier for the rapidly growing smartphone market, with deep ties to Intel, Samsung, and TSMC. It had secured this position by winning the decisive technical battle of the 2000s.

At the start of the millennium, the entire semiconductor industry hit a physical wall. Circuits had been getting steadily smaller for decades by simply switching to shorter wavelengths, but the standard 193-nanometer light (roughly one five-hundredth of the thickness of a human hair) was too blunt to draw smaller circuits.

Nikon tried to solve this by developing a new light source with a smaller wavelength of 157 nanometers. But this shorter wavelength light was absorbed and distorted by standard glass, forcing Nikon to build lenses out of calcium fluoride, a rare, brittle crystal that was expensive to polish and prone to cracking under heat. The industry poured hundreds of millions of dollars into this ‘dry’ lithography path, only to find the manufacturing challenges insurmountable.

ASML’s partnerships helped it avoid this dead end. TSMC researcher Burn Lin had advised them to switch to a technology called immersion lithography. ASML continued to use 193-nanometer light but placed a layer of water between the lens and the silicon wafer. Just as a straw appears bent and magnified when placed in a glass of water, the water in the machine bent the light waves, sharpening the focus and allowing smaller circuits to be printed without needing new lenses.

ASML compounded this advantage by introducing a revolutionary machine architecture called TWINSCAN. In older machines, the light source would sit idle while the machine stopped to measure the surface of the silicon wafer to ensure it was flat. ASML replaced this with a dual-stage system: a massive machine with two tables would measure one wafer in the background while another was being printed simultaneously. This eliminated the dead time in the manufacturing process, allowing chipmakers to produce significantly more chips per hour. By the time Nikon abandoned its 157-nanometer project in 2005, ASML had become the industry standard, with 53.2 percent of the market.

ASML’s machines were so much better than the competition that it could charge nearly twice as much for them: $55 million versus $30 million for the comparable Nikon device.

But even this was not enough. While ASML was beginning to ship prototype EUV machines to IMEC from 2006 onwards, they were so slow and prone to breaking down that they were commercially useless. In 2012, ASML, still reeling from the global financial crisis, was struggling to continue financing its EUV efforts.

In a drastic move – part desperate attempt to keep the company’s research efforts afloat and part strategic bet to win the EUV market once and for all – the ASML leadership launched a co-investment program that sold 23 percent of the company to its three largest customers: Intel, TSMC and Samsung.

The funding also allowed ASML to complete a $2.5 billion acquisition of one of its suppliers, Cymer, which produces lithography light sources. The acquisition allowed ASML to invest in Cymer’s R&D efforts to perfect its soft X-ray light source, which involved hitting fast-moving droplets of tin with such force that they lost electrons, but precisely enough that this did not shed so much debris that it coated the mirrors. They accomplished this by moving from a single pulse to two separate laser pulses: the pre-pulse would shape the droplet and the main pulse would generate the plasma. This improved efficiency and stability.

ASML’s close partnership with TSMC proved especially critical. In 2014, TSMC launched its first chip for Apple, which was now its largest customer and was putting pressure on the chipmaker to produce higher performance chips than its existing machinery was capable of. It had become urgent for ASML to complete a commercial EUV machine.

The two companies worked so closely together that Anthony Yen, the Division Director at TSMC responsible for overseeing EUV development, described them as ‘one team’. ASML and TSMC engineers on the ground worked tirelessly, troubleshooting and iterating until they had reached the necessary throughput: 500 wafers a day for a month.

During this period, the joint team redesigned both the tin-droplet generator and the way the laser hit each droplet. The new setup produced droplets that were about half the original size while still yielding the same ultraviolet energy. Smaller droplets throw off far less debris when vaporized, which slows the rate at which tin builds up on the collector mirror. Because the mirror degrades more slowly, it needs fewer replacements, keeping the machine up and running for longer stretches.

The partnership was a win for ASML, as it was able to work through some of its key engineering and commercialization challenges. It also helped TSMC become an early adopter of the most cutting-edge technology. By 2019, TSMC was ramping up mass production of its seven-nanometer process and the first phones with EUV chips were being sold by the end of the year.

Meanwhile, competitor firms like Nikon, which had never believed as strongly in extreme ultraviolet lithography, effectively gave up. In its 2013 annual report, Nikon noted that its own EUV progress had not proceeded as planned, and it was not mentioned in an annual report again. With ASML pulling ahead on R&D and locking up key customer demand, and with competitors struggling to justify their own R&D spending in the wake of the financial crisis, ASML became the last company standing in the race to commercialize the technology.

The importance of tacit knowledge

Early on, ASML cultivated a culture that was more risk tolerant than other players in the industry. It promoted high-potential talent early and had a track record of retaining key employees for decades. Much of this is a product of its challenging early years. ASML needed the talent of its younger generation to save the company, so it was more willing to promote and empower them quickly.

For example, Martin Van Den Brink joined ASML in 1984. Within 18 months, aged 29, he became one of two people promoted to lead the development of one of the company’s early flagship projects. He carried on working at ASML for his entire career, serving as president and chief technical officer until his retirement in 2024. This practice was far less common among ASML’s Japanese rivals, who were more hierarchical and tended to reward seniority over performance.

Retaining the best workers is especially crucial in an area like photolithography, where a huge amount of tacit knowledge is used to assemble its machines. An ASML engineer once told He Rongming, the founder of Shanghai Micro Electronics Equipment, one of China’s top ASML competitors, that the company wouldn’t be able to replicate ASML’s products even if it had the blueprints. He suggested that ASML’s products reflected ‘decades, if not centuries’ of knowledge and experience. ASML’s Chinese competitors have systematically attempted to hire former ASML engineers, and there is at least one documented case of a former ASML employee unlawfully handing over proprietary information. But none of this appears to have narrowed the gap.

A European giant

ASML is a rare example of a European tech giant. Its success was the result of transatlantic cooperation, not continental parochialism. Had the company not joined a program funded by US chipmakers, Canon and Nikon would likely still dominate a less advanced lithography industry.

Cooperation with other companies was just as important. While vertical integration gave Nikon and Canon total control, it capped their innovation at the limits of their internal resources. In a system exceeding one hundred thousand components, that ceiling proved fatal. ASML’s modular approach allowed it to import cutting-edge physics by acquiring Cymer and investing in Zeiss, while distributing the risk to customers like Intel and TSMC. This strategy created a collective engine that outspent and outpaced every rival attempting to shoulder the burden alone.

This took a great deal of courage. ASML sank billions of dollars into the development and commercialization of EUV technology, with no guarantee that it would ever work. As late as the 2010s, many semiconductor experts doubted that the technology could be successfully commercialized. Now it is the most important technology in the world.

But ASML, and by extension the continent, cannot stand still. As ASML enjoys its place as an indispensable pillar in one of the world’s most important industries, others are working to create a new paradigm in chip technology. Moore’s Law probably doesn’t end here, and in a matter of years, five nanometers won’t be small enough.

Neil Hacker runs partnerships at Isometric. Follow him on Twitter.

On 17th October 1956, Queen Elizabeth II pressed the switch that activated Calder Hall, the world’s first grid-scale nuclear power station. As the country attempted to reassert world leadership after the trauma of World War II, The Times’s correspondent excitedly recalled that:

Today, with a boisterous wind to display the flags – and nearly wreck the marquees – the colourful and almost Wellsian-looking installation deeply stirs the imagination. Nothing like it exists elsewhere. Truly it has been described as a ‘courageous enterprise’; for Calder Hall represents the inauguration of a comprehensive programme of atomic power stations which, in time, will provide Britain with an ample supply of electricity without the use of coal or oil. Therein lies its magic.

By 1965, Britain had built more nuclear power stations than the US, USSR, and France combined. Stations were approved in months, regularly built in under five years, and operated with an impressive safety record. But in 1970, the UK’s lead faltered and has never recovered.

Hinkley Point C, under construction in the southwest of England, is set to be the most expensive nuclear power station ever built anywhere in the world. Each unit of power generated will cost six times more than one produced by a modern South Korean nuclear power station or an early British reactor. By the time the plant comes online, it will have taken at least 13 years to build and will have run £17 billion over budget. Britain last successfully completed a nuclear reactor in 1995.

Most accounts emphasize the importance of technology. The story goes that the UK’s 1965 decision to embrace a convoluted domestic reactor design created a series of white elephants that rendered nuclear power uneconomic in Britain. After privatization, nuclear power was written off by the markets, and the British nuclear industrial and skills base left to shrivel. This view was also widely held at the time. As early as 1972, Sir Arthur Hawkins, the chair of the UK’s powerful state-run energy company and regulator, dubbed the choice ‘an economic catastrophe we must not repeat’.

There is a grain of truth in this story, but it is incomplete. The tide turned against nuclear only after Britain had switched over to proven American technology in the 1980s. Meanwhile, even the clunkiest British reactors were made to operate efficiently once they were placed in the hands of more professional management in the 1990s.

Instead, the story of British nuclear power more closely resembles the story of the British state in the postwar period. At first, Britons appeared to be living in the age of the technocrat. Talented scientists and engineers armed with a near carte blanche from the government to build with minimal consultation seemed unstoppable. But somewhere along the way, this model broke down. It didn’t take incentives seriously, underdelivered, and failed to adapt to changing public expectations. Experts failed, and the public withdrew the blank check they had given them. The story also serves as a valuable corrective to the idea that the world would work better if we simply handed more power to engineers.

Subscribe now

The first wave

In October 1946, in a bad-tempered cabinet committee meeting, British Foreign Secretary Ernest Bevin thundered: ‘We’ve got to have this thing over here whatever it costs. We’ve got to have the bloody Union Jack on top of it’. Bevin was infuriated by the American decision to cut off all nuclear cooperation with its wartime allies, including those which had helped it develop the atomic bomb. Casting aside Treasury reservations, the government concluded that if it could not share a nuclear deterrent with the Americans, the empire on which the sun never set must have its own.

To arm the bomb, Britain needed plutonium. Drawing on the wartime memory of British scientists who had worked at Los Alamos, the UK built the Windscale Piles. These reactors, named because they were stacks of graphite blocks, would serve as the inspiration for the first civilian UK reactor: the Magnox.

A Magnox reactor packed unenriched metallic uranium fuel rods into a graphite core. The design was named after the magnesium alloy cladding that stopped the uranium fuel rods from reacting chemically with the gas. If you pulled out the fuel rods quickly enough, the neutron reactions taking place in the reactor converted the uranium into weapons-grade plutonium, but if left alone, it also generated electricity.

The Magnox program was born in 1953, when the government gave the go-ahead for Calder Hall. In response to the growing secret cost of the British nuclear weapons program and a desire for greater energy security, the government decided to find a civilian use for the technology.

The most striking element of the Magnox program was its speed: between 1956 and 1971, Britain built 26 reactors. Projects were routinely approved within months, and reactors typically achieved grid connection within four to five years of starting construction. This is all the more impressive given that there was no standard Magnox. The Atomic Energy Authority played a significant role in coordinating research and design, as well as in training consortium staff, but as there was no British company with the skills and capacity to build an entire nuclear power station, the plants were built by competing consortia, each to slightly different designs.

The approvals process was straightforward. The Central Electricity Authority, the body responsible for managing Britain’s nationalized energy supply, conducted site studies and consulted local authorities to identify suitable locations. This consultation would factor in space, proximity to a water source, terrain, and existing development plans. Once the authority had selected a site, the minister would publish a notice in The London Gazette and in the local press, as well as notifying anyone who owned or leased land within 300 yards of the site. They could then submit a letter of objection to the Ministry of Power.

If they received significant objections from either locals or a public body, the minister would hold a local planning inquiry. Before 1957, there was no legal requirement to do so, but ministers held them anyway as a means of avoiding local discontent.

The early nuclear planning inquiries were sedate affairs. In 1958, the minister ordered an inquiry into the prospective Trawsfynydd nuclear power station in Snowdonia. The local authority and community were both supportive, but the National Parks Commission was concerned about the impact that the station would have on the scenery. Over three days, one inspector from the Ministry of Power and another from the Ministry of Housing and Local Government heard representations from both sides. Their final report, including appendices, ran to just 50 pages. Just the environmental assessment for Hinkley Point C ran for over 44,000 pages.

After the inquiry, the minister would give consent to build the power station under the 1909 Electric Lighting Act, the standard legal mechanism for approving new electricity generation at the time. The minister would then invoke section 35 of the 1947 Town and Country Planning Act, which allowed the government to grant planning permission for public developments without involving the local authority.

This straightforward approach reflected broad public support. For example, while conservation groups were concerned about the damage Trawsfynydd might do to the landscape in Snowdonia, the local community welcomed the power station. Merionethshire, the county that hosted the site, had lost 20 percent of its population since 1880 due to economic stagnation. Former residents wrote letters expressing their hope of moving back if the power station created new jobs.

This system needed refinement. These early nuclear power stations were run by the UK Atomic Energy Authority, a government body that enjoyed immunity from civil liability and most regulation. The government’s ambition was that as nuclear scaled, it should be treated like any other energy source, rather than a special technology to be built by a secretive government body.

Against this backdrop, the UK introduced its first licensing system for nuclear power in 1959 by legislation. The Minister of Power would be responsible for issuing licenses and setting conditions. The government also created the Nuclear Installations Inspectorate within the ministry, initially with a team of just 13 inspectors. By comparison, the modern Office for Nuclear Regulation has over 400 specialist technical staff.

In 1960, the ministry granted the first prospective license under the new scheme to Bradwell, a Magnox station in Essex. The license was a modest affair with just seven conditions attached to it. These were light touch by modern standards and included obtaining the consent of the minister before fueling the reactor, keeping records, and testing equipment and materials.

Trouble ahead

The Magnox reactors served Britain well. Wylfa, the last Magnox standing, only came offline in 2015. Reconstructing Magnox lifecycle costs is tricky. But they appear to have been cheap. One reconstruction, which factored in numbers from early Magnox decommissioning, placed the lifecycle costs of electricity from Magnox reactors at eight pence per kilowatt hour in modern money. This would be 5.3 pence per kilowatt hour cheaper than electricity from Hinkley Point C.

The design undoubtedly had its flaws. It relied on natural uranium, which is good for producing weapons-grade plutonium, but not for power generation. Britain initially had little choice but to go down this route as it lacked domestic enrichment capabilities, and the US refused to export enriched uranium. At the time, enrichment was difficult, and involved converting uranium into a gas and forcing it through thousands of porous metal barriers. By the time Britain’s first specialist enrichment facility kicked into gear in the mid-1950s, the country had already settled on a design for civil reactors.

The usable fuel content in natural uranium is quickly burned up, forcing frequent refuelling. The fuel’s Magnox coating corroded rapidly, losing strength when temperatures in the reactor rose above 400 degrees Celsius. These limits meant the reactors had to run cooler, producing lower pressure steam and wasting more of the heat generated by the fuel. As a result, the Magnox reactors generated electricity about one third less efficiently than American light water reactors.

By the early 1960s, the government was searching for a new design. An intense bureaucratic battle played out behind the scenes. The Central Electricity Generating Board was predisposed towards an inexpensive American light water reactor design, as the US had resumed nuclear cooperation with the UK a few years earlier. Meanwhile, the Atomic Energy Authority feared that a switch away from gas-cooled, graphite designs would deprive them of valuable R&D and design work.

Instead, they championed the Advanced Gas-cooled Reactor (AGR), an upgraded version of the Magnox, which the Atomic Energy Authority had prototyped a decade earlier. The AGR used enriched instead of natural uranium, and each fuel rod was sealed in stainless steel, which could handle much higher temperatures. This allowed the plant to run at twice the temperature of the Magnox and, theoretically, higher levels of thermal efficiency than a light water reactor. This was scientifically elegant, but untested at scale. Choosing the design would also drive orders towards British firms.

In 1965, the Ministry of Power decided on the tender for Dungeness B, the second reactor to be built on the Dungeness headland in the south of Kent. In the end, through a combination of sharp bureaucratic elbow work and low skulduggery, it went ahead with a twin-reactor AGR station.

Subscribe now

The original sin of British nuclear?

Amid stiff competition, Dungeness B is arguably the worst British infrastructure project of the modern era.

Atomic Power Constructions Limited, which won the tender to build Dungeness B, was a borderline moribund company. It had submitted a token bid with no expectation of winning, as a way of signaling to the Central Electricity Generating Board that the company was still alive and hopeful of future work. The Atomic Energy Authority covertly helped the company produce its supporting technical documentation. It underbid significantly on price and committed to an ambitious four-year schedule, promising to beat Magnox timelines while building a significantly larger and more complex first-of-a-kind design.

As one former employee observed later: ‘Nobody was more surprised than the staff of APC when their offer for an AGR secured acceptance.’ Major technical problems and shoddy work blighted almost every element of the plant. Engineers had to redesign the boilers after they couldn’t fit them into the available space. A key supplier of gas circulators went bankrupt, which forced the project team to scramble for replacements. On site, workers welded the steel poorly and installed the wrong materials in pipe and boiler supports, while the Nuclear Installations Inspectorate ordered extensive redesigns. Atomic Power Constructions went bust four years into the project and other companies had to step in and rescue the project.

Even so, work had already started on further AGR stations at Hinkley Point B, Hunterston, Hartlepool, and Heysham, split across different consortia. Having different companies build different reactors to a similar specification had worked well enough for the small and comparatively straightforward Magnox units, but it underperformed for the novel and more complex AGR. Different teams ran into similar engineering issues with the reactor and boiler and invested time resolving them independently. All ran behind schedule.

By 1974, nine years after the start of the program, Britain still had no operational AGRs, while costs had escalated by an average of 50 percent. The Central Electricity Generating Board concluded that ‘no further orders for AGRs can be contemplated in the near future’. For three years, officials contemplated another design switch.

By the end of the decade, even the AGR’s most ardent bureaucratic defenders had given up the ghost and advised dropping the design in favor of the American pressurized water reactor (PWR). But Tony Benn, then Secretary of State for Energy, ordered another two twin-unit AGR stations in 1978. In what reads like a parody of industrial strategy, Benn justified the move on the grounds that it would preserve the jobs of turbine and boiler makers.

The first reactor at Dungeness came online in 1983, 13 years behind schedule, four times over budget, and operating significantly below capacity.

In many accounts, the selection of the AGR was the original sin of British nuclear: a bad technology choice that tainted everything that followed. By the mid-1970s, a steady chorus of voices was already labeling the program a costly mistake and drawing unflattering parallels with Concorde, another high-profile, state-led, technically ambitious program that was insulated from market forces. There is clearly some merit to this view.

But simply blaming the AGR for the failure of UK nuclear seems unsatisfactory. While the early builds were disastrous, the last two stations that Benn ordered, Heysham 2 and Torness, were completed in eight years. Heysham was constructed within five percent of its original cost estimate. In 1973, Britain abandoned the consortium model, merging the different construction companies into the National Nuclear Corporation. This improved coordination and meant that, unlike earlier AGR builds, these latter stations shared a common design.

In short, the AGR wasn’t perfect, but by itself, the technology choice didn’t have to be fatal. And it is striking how little impact the shortcomings of the AGR program had on either public or elite opinion about nuclear power. As late as 1978, when the Dungeness B saga was widely known, 57 percent of the British public and every major political party continued to support nuclear power. The debate was confined to the technocrats, who were focused on design choice as opposed to the merits of pursuing a program at all.

The tipping point

Ironically, British nuclear’s downfall really came after the government decided to back the right technology. When Margaret Thatcher came to power in 1979, her government saw nuclear power as a vehicle for breaking the power of the coal mining unions that had held successive governments to ransom.

Sharing none of her predecessors’ desire to preserve technological independence from America, and keen to adopt a standardized design, the government embraced the PWR. While the PWR was less fuel efficient than the AGR on paper, it was simpler. By using the water for both cooling and moderation, the PWR combines both functions into one sealed, high-pressure circuit, avoiding the AGR’s separate graphite core and gas coolant system. The Central Electricity Generating Board agreed, arguing that the PWR had ‘substantially lower capital costs per unit of electricity than AGR’ and was ‘proven mainstream worldwide technology with some 150 reactors in operation worldwide, compared to 5 AGR stations in the UK’.

Unfortunately, in 1979, a PWR plant was involved in what was then the worst ever civilian nuclear accident. The meltdown at the Three Mile Island Generating Station in Pennsylvania, despite causing no deaths or illnesses, catalyzed opposition to nuclear power around the world. A plurality of the British public remained supportive, but only narrowly so. The old approach of debating nuclear policy among technocrats with little public input appeared to be reaching its end.

To allay these concerns, the government decided to pursue a policy of maximum transparency. Announcing the government’s planned PWR program in December 1979, the Secretary of State for Energy David Howell spoke of ‘the Government’s determination to ensure that full information is in the hands of Parliament and of the public before and as we make decisions to expand our nuclear capacity’. They decided to subject Sizewell B, the first PWR, to a full-scale inquiry, intending it to reassure the public about its safety.

Sir Frank Layfield, a distinguished planning lawyer, was appointed to lead the inquiry, which started work in 1981. The government gave Sir Frank unusually broad terms of reference. As well as the impact of the development on the local environment and amenities, the inquiry would cover the safety features of the design itself (something normally confined to the licensing process) as well as the need for a power station at all and the government’s long-term energy policy.

The resulting inquiry sat for 340 days. Instead of reassuring the public about the safety of Sizewell’s design, the inquiry descended into a public debate about government energy policy in general. Weeks were dedicated to military uses of nuclear technology, the ethics of importing uranium from Namibia (then occupied by apartheid South Africa), the impact of uranium mining on Australian aboriginal land rights, and other countries’ experimental reactor designs.

The technocrats, unprepared for this degree of scrutiny, acquitted themselves poorly in the public gaze. The Central Electricity Generating Board had not completed crucial safety documentation before the start of the inquiry and repeatedly filed addenda. The Nuclear Installations Inspectorate gave the impression of being subordinate and entirely reactive to the Board.

The inquiry was also dissatisfied with the quality of the evidence the Board supplied on construction timelines and the economics of nuclear power versus fossil fuels, dubbing it ‘insufficient for a proper understanding of either the approach or the assumptions adopted’.

Subscribe now

Overreach

This might not have mattered too much if it had been the sole example of bureaucratic arrogance. During the 1970s, against a backdrop of growing concern about the environment, British Nuclear Fuels Limited began negotiating contracts to reprocess spent nuclear fuel from Japan at Sellafield, the nuclear site in the north of England that hosted Calder Hall and the Windscale Piles. While the plan had openly been discussed and written about in the nuclear trade press, it largely escaped public and political attention, until a 1975 Daily Mirror front page warned of the ‘Plan to Make Britain World’s Nuclear Dustbin’.

The ensuing public outcry prompted the Royal Commission on Environmental Pollution to study the question of nuclear’s impact on the environment. Its 1976 report recommended that: ‘There should be no commitment to a large programme of nuclear fission power until it has been demonstrated beyond reasonable doubt that a method exists to ensure the safe containment of longlived, highly radioactive waste for the indefinite future’.

Amid this outcry, industry was largely silent. In the eyes of both the officials and engineers, nuclear waste wasn’t a real problem, so they refused to take the public debate seriously. They assumed that nuclear waste would be recycled and used to fuel fast breeder reactors, which would convert the uranium in spent fuel into new plutonium for energy generation. As a result, waste was left to pile up in tanks on the basis that it would be resolved later.

The fast breeder reactor proved uneconomical and the British program never progressed beyond a small demonstrator plant. It also had another undesirable side effect. As the fast breeder reactor produced plutonium as a byproduct, it was seized on by opponents of nuclear power as proof that the civilian program was supporting military efforts. Frank Barnaby, a former nuclear scientist at the Atomic Energy Authority, had written Man and the Atom, a passionately pro-nuclear book in 1971. The book went as far as advocating nuclear explosions to dig canals. In 1975, after the fast breeder reactor program was unveiled, he published a new edition of the book. While much of the text was reproduced verbatim, he drew much more pessimistic conclusions and advocated for non-proliferation.

Disposal facilities in the UK met significant backlash in more or less every community where they were proposed, while the Sellafield site became a focal point for protests. In 1978, a year after its formation, Greenpeace began to publicize the dumping of nuclear waste at sea by the Atomic Energy Authority. After five years of embarrassing publicity led to the threat of a boycott by ship and transport workers, the government stopped the practice. The failure to anticipate the public outcry is arguably all the more confusing, given that nuclear waste had sparked outbreaks of local resistance as far back as the 1950s. In 1954, two years before Calder Hall opened, the Atomic Energy Authority had proposed to dump waste in the Forest of Dean, only to be defeated by local residents invoking obscure thirteenth-century land rights.

This persistent failure may have been a product of the industry’s culture. The founders of the Atomic Energy Authority and godfathers of the British nuclear industry were figures like Christopher Hinton, John Cockcroft, William Penney, and Edwin Plowden. They were products of World War II nuclear research, where they enjoyed sweeping authority to pursue a high-level mission with little interference or need for consultation. This mindset suffused the industry they helped to build.

In their defence, the technocrats were struggling to keep up beyond nuclear. Public opposition to megaprojects reached boiling point during the 1970s.

Between 1965 and 1973, the newly created Greater London Council embarked on a plan to build four motorways, stretching for 478 miles, through many of London’s most prized historic neighborhoods. Before the Ringways scheme, the planners almost always got their own way. Opposition to infrastructure of any kind had historically been localized, never widening into a broader movement. With the Ringways, the planners tried to fight too many people at once, lost, and demonstrated that public opposition to megaprojects could work. Similar revolts against high modernist planning were unfolding internationally at the same time. The most famous example was in New York, where Robert Moses’s highway and urban renewal schemes faced neighborhood campaigns led by figures like Jane Jacobs.

Sometimes the technocrats could open the door for the opposition themselves. In 1971, a commission planning a third airport for London selected the village of Cublington in Buckinghamshire. One member of the commission objected on environmental grounds, legitimizing popular opposition. After a political firestorm, the airport was moved to a more expensive site in the sandbanks off the Essex coast, in the Thames Estuary, eventually rendering it economically unviable. The project was canceled in 1974, and various attempts to revive it, most recently in 2016, have flamed out.

Another chink in the armor was visible with the new towns program. The government bulldozed through the opposition of local residents and even court injunctions in the 1940s to designate new towns like Stevenage, Cumbernauld, and Peterlee. These were idealistically planned to relieve overcrowding in bombed-out cities like London and Glasgow. Architecturally, the experimental designs often failed, producing alienating concrete jungles. The last new town project to be pursued to full completion was Milton Keynes in 1967.

Subscribe now

The chill

While the seeds of the backlash were being sown, nuclear licensing was becoming more conservative. In 1974, the government had consolidated workplace health and safety regulation into a single unified body, including sectors like nuclear power. The new Health and Safety Executive, which the Nuclear Installations Inspectorate joined, mandated that risk in the workplace be reduced ‘as low as reasonably practicable’ (ALARP).

Crucially, the Health and Safety Executive was a nondepartmental body. The Ministry of Power had the clear objective of building new power stations. It wasn’t in the minister’s interests to rush approvals and build unsafe stations, but nor was it in their interests to endlessly postpone. The same dynamic is visible elsewhere: France, which undertook a vast nuclear construction program in the 1970s and 1980s, did not have an independent nuclear regulator until 2006, while in South Korea, which builds the developed world’s cheapest reactors, the prime minister’s office is heavily involved in nuclear regulation. When Britain made the licensing process independent, this link was broken. There was no longer an actor in the regulatory process incentivized to say yes.

The influence of ALARP is clear on Sizewell B. While theoretically based on an existing US design, the Central Electricity Generating Board added an extra containment dome, two additional backup systems to complement the two already in the design, and built in greater earthquake protection. This theoretically enhanced safety, but at a high cost. New guidelines, published in 1992, stated that safety measures must be taken until the cost is ‘grossly disproportionate’ to the risk averted. This meant that, depending on the risk, measures could be between two and ten times more expensive than the equivalent harm averted.

The combination of a less deferential population and a system of safety-focused but cost-indifferent regulators increasingly collided with an undiminished bureaucratic appetite for megaprojects. By ignoring concerns about waste, the nuclear industry alienated the public. By being seen to fudge the figures on nuclear economics and attempt to strongarm regulators, it offended politicians. Amid mounting skepticism, the industry was fast running out of friends, just when it began to really need them.

The premature death of UK nuclear?

The construction of Sizewell B began in 1988 and was completed in 1995, making it the last new reactor to be completed in the UK. The project cost approximately £6 billion in 2025 money for a 1.2-gigawatt reactor, making it roughly three times cheaper than Hinkley Point C per gigawatt generated. It was built in roughly the time it took to build a Magnox reactor, despite generating three times as much electricity and facing much stricter safety requirements. Despite this apparent success, the writing was on the wall for the industry.

The government scrapped plans for three more PWRs, keeping only the plans for Hinkley Point C. The Central Electricity Generating Board applied for permission to start work on the project in 1987, triggering a 14-month inquiry that heard from 600 witnesses and navigated 22,000 objections. The final report, completed in 1990, recommended approval, but suggested a final decision be postponed until a full government review of nuclear policy had taken place.

In 1995, the government formally announced that it would no longer fund the construction of nuclear power and that there is ‘no case for Government intervention to distort the electricity market by providing finance or guarantees for one form of generation over another’. It acknowledged that ‘private finance is unlikely to be available at present for new nuclear construction’.

This lack of enthusiasm stemmed from the industry’s most public defeat. In 1987, the Conservatives had won reelection on a platform that included electricity privatization. As investors began to scrutinize the operating record and finances of the AGRs, they came away underwhelmed.

More importantly, they didn’t want to take on responsibility for the cost of decommissioning the stations once they had reached the end of their lives. Decommissioning, which is regulated exceptionally conservatively in Britain, involves removing spent fuel from the reactor and cooling system, dismantling the plant, and decontaminating the site. The Central Electricity Generating Board had valued the decommissioning liabilities at £10.4 billion (in 2025 prices), but when auditors came to scrutinize the numbers, their estimate ballooned to between £22 and £36 billion. Meanwhile, without life extension work, the continued operation of the Magnoxes would generate less in revenue than their decommissioning liabilities.

The government concluded that the Central Electricity Generating Board had been deliberately burying the true cost of the nuclear program for years, subsidizing it with the revenue from coal and oil power stations. In the autumn of 1989, on the eve of privatization and having failed to find private sector buyers, the government decided to keep nuclear in public ownership.

Subscribe now

Nationalization with Thatcherite characteristics

This exclusion from privatization was a public humiliation for the UK’s nuclear industry, but also marked the beginning of a turnaround. With the Central Electricity Generating Board dismantled by privatization, the government shifted the AGRs and Sizewell B into two state-owned companies: Nuclear Electric and Scottish Nuclear Limited.

The sheer scale of the UK’s nationalized system had made it easy to bury losses. Individual stations were not profit centers, so their management had little incentive to drive efficiency improvements. Meanwhile, status lay with engineers and scientists, rather than the managers of individual power stations: the manager of a facility like Torness could not authorize payments above £14,000 in today’s money without consulting Whitehall. The Central Electricity Generating Board, which was keen to buy the PWR, was uninterested in improving the economics of the AGR and thus undermined their own business case.

With nowhere to hide, and the old technocrats replaced with professional managers, performance improved. Managers studied higher-performing American and European reactors and embedded better preventive maintenance. Nuclear Electric ran a quality improvement process that mapped every gap between theoretical maximum AGR output and the actual number of megawatts generated. Engineers who identified cost savings or output improvements received bonuses.

Between 1990 and 1996, the English AGR fleet went from being operational 47 percent of the time to 76 percent of the time. This was the equivalent of adding two new power stations. In 1990–91, Nuclear Electric’s operating loss stood at £1.1 billion. By 1995–96, before Sizewell B had started commercial operation, the AGR fleet was profitable for the first time.

In 1996, the government merged Nuclear Electric and Scottish Nuclear Limited into British Energy and privatized it through a stock market flotation. Over the next few years, it paid steady dividends, cut debt, and was crowned ‘the darling of the utilities’ by the Financial Times. IPO investors were sitting on 355 percent gains by the start of 1999. It even began to expand into the US. But hubris struck. Under pressure from the City, the board handed back almost £1 billion to shareholders in 1999 through dividends and buybacks, stripping away the firm’s financial buffer.

British Energy was then hit by a turn in the market. Privatization helped enable the ‘dash for gas’ in the UK as private companies rapidly built cheap gas generators; gas-generated electricity jumped from 1 gigawatt in 1990 to 23 gigawatts in 2000. Initially, the price was set by the most expensive generator needed to meet the demand in a centralized mechanism called the Pool, which slowed the impact of cheap gas-fired plants. But in 2001, the government abolished the Pool, leading generators to strike direct contracts with electricity suppliers and large customers, causing prices to collapse to their lowest levels since privatization.

British Energy was already locked into expensive fuel reprocessing and waste management contracts. Even with the improved efficiency, operating and maintaining an AGR was expensive. In September 2002, the government stepped in with a £410 million emergency loan to keep the reactors running. As part of a restructure, the government took on responsibility for decommissioning and waste costs, while shareholders were wiped out. In 2009, the government arranged a sale of British Energy to France’s state energy company EDF.

Unlearning nuclear

By the late 2000s, concerns about climate change and rising gas prices led the government to return to nuclear power as an option. EDF revived the idea of building a third nuclear power station at Hinkley Point. Instead of investing directly, the government committed to pay a fixed, inflation-protected price of £92.50 per megawatt-hour generated in 2012 money, or £133 in 2025 pounds. This was significantly higher than the wholesale price has ever been since the electricity market was created, except during the crisis caused by Russia’s invasion of Ukraine.

For Sizewell C, the next new nuclear project, the government abandoned this model on value for money grounds. Instead, it had to finance the cost of development upfront and take on 45 percent of the investment itself.

The process for agreeing financing built in significant extra time. The negotiations over Hinkley Point C’s financing took four years, while designing and agreeing the funding mechanism for Sizewell C, Britain’s next planned nuclear project, took six.

By this stage, Britain had the most conservative regulatory regime in the world for nuclear power, excluding the countries where the technology is banned. This system results in designs considered safe in other countries having to go through extensive UK-specific modifications.

For example, Chinese, French, and Finnish regulators judged that the design’s two separate digital instrumentation and control systems offered sufficient redundancy in the event of a software fault. The UK regulator, by contrast, required that an entirely new analog system be designed from scratch. The system took 12 years to develop, approve, and build, and required an additional 76 equipment cabinets in each of the power station’s two units, along with a redesign of a number of auxiliary buildings. The 7,000 changes to the design meant Hinkley Point C required 35 percent more steel and 25 percent more concrete as a result of UK-specific modifications.

Since Sizewell B, the UK’s planning and environmental laws have tightened significantly. During the golden age of British nuclear, the developer needed to secure a single approval to start building. By contrast, modern developers have to assemble reams of parallel consents from different public bodies, operating to their own rules and timetables. The same vetocracy that has slowed the construction of roads, airport runways and tunnels has clamped down on nuclear with a vengeance.

These rules would always be burdensome, but the financing model for nuclear does not incentivize cost control. For example, the model for Sizewell C means that the cost of overruns is ultimately borne by a mixture of the billpayer and the taxpayer, not developers. This also means that developers have little incentive to challenge overly cautious rulings from regulators. There is a formal appeals process against nuclear site licensing decisions, which goes almost entirely unused.

Subscribe now

The twilight of the technocrats

When Queen Elizabeth II flicked the switch at Calder Hall, an electronic dial immediately lit up showing the number of kilowatts the station was generating. Years later, one of the engineers discovered that this was powered by a man turning a handle. Seemingly impressive, but opaque and held together by improvised solutions, this set the tone for the British nuclear industry.

Every successful nuclear buildout, whether it is the Magnox reactors, France in the 1980s, or China today, shares a number of ingredients in common. They all had unambiguous political backing, regulators incentivized to approve promptly, predictable demand, and a supportive or indifferent public. Between 1965 and 1995, every one of those pillars crumbled in Britain. The scientists and engineers who ran the program, and the political class that provided such weak oversight, were at times the victims of bad luck but largely had themselves to blame. Following the revival of interest in nuclear power in the 2000s, the government, with the buy-in of the private sector, tacitly accepted that it would always be uneconomical, giving rise to a self-fulfilling prophecy.

At the start of 2025, the UK government commissioned an independent taskforce to provide recommendations on how to reverse rising costs and lengthening timelines. The taskforce’s report, which the government accepted in full, concluded that the sector is blighted by ‘a culture of resignation among dutyholders and regulators who simply accept that nuclear projects must be slow and expensive’. Adopting the taskforce’s more radical proposals would go some way to restoring the ingredients for nuclear success in Britain.

Alex Chalmers is a staff writer at the Cosmos Institute. You can follow him on Twitter.

Egg freezing is much more effective than most people think. Articles in major publications like The New York Times (‘Sobering study shows challenges of egg freezing’) and Vox (‘The failed promise of egg freezing’) have reported that only about two fifths of women will be able to successfully have children from their frozen eggs. Articles about adjacent topics, often criticizing companies for offering egg freezing as a perk, give a needlessly pessimistic tone: ‘the odds are stacked heavily against you.’ (Glamour magazine); ‘Success rates for frozen eggs vary widely, but rarely go above 30 per cent’ (Financial Times); ‘Despite the growing popularity of egg freezing for women who want kids eventually but not right now, doctors don’t actually recommend the procedure for this purpose’ (Business Insider).

These articles usually reference a study from 2022 that finds that only 39 percent of patients had a baby, but what this coverage misses is that the average age of the women was 38 when they froze their eggs. Their fertility had already begun a precipitous decline. But declining fertility in women is largely about egg, not womb, aging, meaning that a woman who freezes her eggs in her twenties will have roughly the same chances of successful IVF in her forties. The oldest woman ever to give birth using a young donor egg was 74.

Anti-egg freezing media has real life consequences. Women, in general, do not know how successful egg freezing is. Even the women who undergo egg freezing don’t know quite how good it is. Even medical students and doctors don’t know.

Rather than freezing their eggs in their twenties, when it would be optimal, women often freeze their eggs as a last resort. The rate of egg freezing for women aged 18–34 is the same as that for women between 43 and 44. This is despite the fact that women aged 43 to 44 who use their own eggs have a success rate of less than ten percent even after multiple rounds of IVF, versus about 85 percent for women under 35.

Ultimately, the choice of whether to freeze eggs is personal and will depend on finances, life plans, relationship status, and career choices. Like any medical procedure, it is not guaranteed to succeed. Some women are infertile or just very unlucky. But it is possible for most women and much easier than the press would have you believe, as long as it is done in time.

We are both women in our late twenties. We are freezing our eggs to make sure we can have the number of children we want whatever happens in our lives and careers. One of us is a biologist (Ruxandra) and the other is an economist (Luzia). We have teamed up to demystify a process that is both emotionally charged and improperly covered, and explain why more young women should consider freezing their eggs too.

Subscribe now

Why fertility falls with age

Why does the ability to get pregnant decline with age? The conventional wisdom is that fertility is relatively constant and then drops sharply after 35. Accurately measuring this is challenging as many studies use small samples or adversely selected populations, like those undergoing fertility treatment, but recent research shows that reality is more complicated than the conventional wisdom might suggest.

One issue is the confusion between fertility and fecundability. Fertility refers to the observed number of births per woman in a population over a given period, while fecundability is the monthly probability that a woman who has regular unprotected sex will conceive a viable pregnancy. Fecundability reflects the biological capacity for conception, whereas fertility reflects the actual number of children a woman has chosen to have.

Historically, researchers could observe only fertility. One of the most consequential studies in this regard comes from 1986, when researchers analyzed a mix of historical populations (including seventeenth-century Genevan bourgeoisie and twentieth-century Iranian peasants) and produced an age-fertility curve that became the foundation for modern beliefs about age-related reproductive decline. Their figure showed a concave pattern, in which fertility was stable through a woman’s twenties and early thirties before dropping sharply after the age of 35.

However, this fertility data is a poor proxy for fecundability. Many of the younger women in these historical populations were biologically incapable of conceiving, either because they were pregnant or because they had recently given birth. Including these women in the denominator artificially depressed fecundability estimates for younger women. Older women, who were less likely to be pregnant or postpartum, did not suffer the same bias.

Modern data shows that the cliff is a statistical illusion. Once you restrict the denominator to women who can actually conceive, fecundity declines roughly linearly from the early twenties onward, with no abrupt mid-thirties break. But it does decline. This decline in fertility with age is underpinned by two factors: a decrease in the quantity of eggs and a decrease in the quality of eggs.

Running out of eggs

Women are born with all the eggs they will ever have, each sitting in a protective sac called a follicle. Most of the eggs are immature, kept in the spare pool called the ovarian reserve. Only a small number of them grow to become antral follicles. These antral follicles are the eggs that the body will mature and release during ovulation. They will either be fertilized or removed in a menstrual cycle. They are also the eggs that can be recovered in an egg freezing cycle.

Women can be born with anywhere from two to six million eggs in their ovarian reserve, but most will ovulate only three to four hundred over a lifetime. The vast majority of primordial follicles are instead destroyed through a quality control filter that eliminates follicles whose eggs or surrounding cells are metabolically stressed, damaged, or poorly supported.

This steady attrition drastically reduces the ovarian reserve over time. By the time a woman hits menopause, only a thousand of the original two to six million remain. As the ovarian reserve declines, the antral follicle count declines almost proportionally. There is a strong correlation between antral follicle count and age. This relationship helps explain why IVF success rates decrease with age.

Egg quality

Egg quality declines with age too. As she ages, a woman’s eggs lose their ability to maintain the correct number of chromosomes.

Unlike most body cells, which routinely divide and make new proteins, eggs spend decades in a hibernation-like state that lasts until they are ovulated. Protein turnover is minimal during this period. Although eggs live for a remarkably long time compared to other cells, this hibernation does not stop all the effects of aging: oxidation, DNA damage, epigenetic drift and mitochondrial dysfunction all still happen. As a result, the proteins that run cell division become damaged or depleted, and the molecular pieces needed to guarantee the right number of chromosomes start to fail.

The cohesin complex is a ring of proteins that clamps each pair of chromosomes together until the moment they should separate. When the cohesin complex ages, the clamp degrades, potentially leading to a premature separation of the chromosomes.

Centromeres, the protein docking sites on chromosomes, also lose structural integrity with age. Normally, these form sturdy attachments to spindle microtubules, the protein cables that move chromosomes during cell division. When they weaken, chromosomes are more likely to attach improperly or be pulled in the wrong direction.

Together, weakened cohesion and faulty attachments cause chromosomes to split unevenly. This means more eggs end up aneuploid (with too many or too few chromosomes) and a smaller share remain euploid (with the normal set). Aneuploid eggs give rise to aneuploid embryos, which usually fail to implant, or result in early miscarriage. In the rare cases that aneuploid embryos can be carried to term, they grow into people with severe developmental disorders such as Down syndrome (three copies of chromosome 21, called trisomy 21), Patau syndrome (trisomy 13), and Edwards syndrome (trisomy 18), all of which increase in incidence with maternal age, sharply after 28.

The egg freezing and IVF journey

Before starting an IVF cycle, women undergo a series of medical evaluations, including blood tests and hormone measurements. The baseline transvaginal ultrasound measures the antral follicle count, the best estimate of how many eggs can be retrieved, and in turn the best indicator of IVF success that we have, given a woman’s age and health.

The actual IVF cycle begins with a drug regimen that induces multiple follicles to mature over 10–12 days. During this time, the patient undergoes ultrasound monitoring every two or three days to track follicle growth. Once the follicles reach the desired size, they are collected through a minor surgery. Roughly four fifths of all follicles removed yield eggs, and roughly four fifths of those eggs are suitable for fertilization. The suitable eggs are then inseminated and cultured for about a week before they are ready for transfer. This stage suffers from high attrition, and only about half of the fertilized eggs complete it. Of these, one is typically selected for transfer to the uterus in the hope of achieving a pregnancy and live birth.

​​Egg or embryo freezing follows the same course, except the process either stops after the eggs have been retrieved or after embryos have formed. The collected eggs or embryos are instead frozen and stored until the patient is ready to get pregnant. Historically, this relied on slow freezing, a gradual cooling method that took several hours and often formed damaging ice crystals inside the cells. Since the mid-2000s, a rapid flash freezing technique, which cools eggs to minus 196 degrees Celsius almost instantly, has been used instead, preventing ice crystal formation and dramatically improving survival rates. Once frozen, the biological clock for these gametes or embryos is effectively stopped. Large-scale studies show that frozen eggs almost always retain their quality and chromosomal integrity, and thawed eggs that survive defrosting perform as well as fresh eggs of the same biological age.

How many eggs can you expect per cycle?

Women between 18 and 34 in developed countries undergoing embryo freezing cycles collect a median of 14 eggs per cycle. There is a lot of natural variation. A cycle can yield anything from zero to more than 50 eggs, even among women at peak fertility. But since roughly 80 percent of a woman’s antral follicles are retrieved as eggs, knowing how many antral follicles a woman has allows us to predict how many eggs a cycle will yield accurately.

This number is whittled down by the embryo stage. As in normal IVF, half of fertilized eggs become suitable for transfer, and two thirds of these have the right number of chromosomes. Ultimately, for the median egg yield of 14 eggs you can expect 2.9 healthy frozen embryos on average, or to put it another way: young women who go through even one cycle of embryo freezing can often freeze enough euploid embryos to have a majority chance of a live birth.

The numbers are even more optimistic for women in their twenties. Data on egg freezing is often combined across the 18–34 age range, which can lead to an underestimate of the number of eggs retrieved for younger women in that group. The largest study on egg donation programs, carried out on 9,539 women with a median age of 26, reported a median of 20 retrieved eggs and six usable embryos.

Subscribe now

The practicalities of egg freezing

Having outlined how egg freezing works and the probabilities that determine its success, we now turn to the main practical considerations that shape how women pursue it: cost, location, and clinic choice

Many women assume egg freezing is prohibitively expensive because they are exposed to American prices ($17,000 per round) or, to a lesser extent, the UK (where it costs around $8,500). Egg freezing at a Spanish clinic uses the same state-of-the-art techniques and equipment, but costs only around $5,000. Spain is internationally recognized for its excellent IVF and has been a pioneer in adopting innovative reproductive technologies. For example, IVI Valencia was among the first clinics to adopt vitrification for egg freezing and published research on its success rates, helping to establish the technique in everyday fertility practice.

The full egg-freezing process takes about two to three weeks, but many Spanish clinics partner with clinics in the UK and other countries so that all pre-retrieval monitoring can be done locally. In that case, you need to travel to Spain only for the egg retrieval, requiring just a four-day stay, and then again for IVF some years later. For every year you wait, you will likely save several hundred dollars, as storing frozen eggs costs about $500–$1,000 in the US and $200–$500 in Spain. At worst, the price difference pays for two trips to Spain. At best, you will save several thousand dollars on top.

A second major issue is clinic quality, which varies dramatically. For example, low oxygen levels are important in embryo culture because early embryos are not adapted to the high oxygen concentration found in room air. In the body, mammal embryos develop in the fallopian tube and uterus at oxygen concentrations of around two to eight percent, far lower than the 21 percent oxygen of atmospheric air. Yet many clinics still fail to use natural oxygen levels in their embryo culture chambers.

These differences matter: success rates are not uniform across the industry. The top clinics achieve success rates above 60 percent per embryo transfer, more than double those of the lowest-performing clinics, which are closer to 30 percent. Verifying your clinic’s success rate is extremely important: we recommend the CDC website in the US; the UK fertility regulator’s Choose your clinic tool; and the Spanish Reproduccion Asistida ORG.

Clinics do not all use the same metrics and the variation is often confusing. The three most meaningful metrics are success per embryo transfer, which measures the number of live births once an embryo is transferred, success per IVF cycle, which measures live births from the start of an egg retrieval cycle, and frozen egg survival rate, which is the proportion of eggs that survive thawing years later.

We recommend relying on statistics from official agencies rather than the headline numbers advertised by clinics themselves. Agencies across countries report success per embryo transfer and per IVF cycle. No regulatory body requires fertility clinics to report frozen egg survival rates, despite this being a fundamental determinant of whether egg freezing ultimately pays off. While overall clinic quality is correlated with egg-freezing success, patients should look for more specific indicators. Positive signs include clinics that have published peer-reviewed research on their egg-freezing programs, clinics with large and well-established egg donation programs, and clinics with long-standing experience in egg freezing.

Should you freeze your eggs?

Egg freezing’s problem today is opacity, not ineffectiveness. We are freezing our eggs. If you are a woman in your twenties or early thirties who can afford it, and think you are very likely to want to have children, but are not sure exactly when, then we think you should freeze yours too. If you face expensive British or American prices, go on holiday to Spain, where it is equally state of the art but many times cheaper. If you definitely do not want to have children, or want only one and have a partner lined up already, then you may not need to bother.

To achieve a strong chance of a future live birth, we recommend freezing approximately 20 eggs in your late twenties or early thirties. This estimate is based on interpolating across the largest available studies, though the literature is not fully consistent. For example, a 2016 study from Spain found that for women under 35, the cumulative live birth rate plateaued after about 15 frozen eggs, reaching 85.2 percent, with a 95 percent confidence interval between 60.5 and 100 percent. Other studies suggest a slightly higher cutoff, closer to 20 eggs. When donor eggs are used, where the donors are selected for high fertility and health, the cumulative success rates reach almost 100 percent.

Whatever you do, do not listen to the articles that say that egg and embryo freezing do not work. Not only do they work, but they work almost as well as the best way to guarantee your fertility, which is to be young when you try to conceive. Egg freezing is the best way to give women the freedom to control when they have children, without having to worry they will not be able to.

Ruxandra Teslo leads the Clinical Trial Abundance Initiative. You can follow her on Twitter.

Luzia Bruckamp is an economics PhD student at the London School of Economics. You can follow her on Twitter.

Read their guide to all the other practical details of egg freezing here.

Japan is the land of the train. 28 percent of passenger kilometers in Japan are travelled by rail, more than anywhere else in the developed world. France achieves 10 percent, Germany 6.4 percent, and the United States just 0.25 percent. Travel in Japan is over a hundred times more likely to be by rail than travel in the United States.

Japan’s vast railway network is divided between dozens of companies, nearly all of them private. The largest of these, JR East, carries more passengers than the entire railway system of every country other than China and India. Each year, JR East carries four times as many passengers as the whole British railway system, even though it has fewer kilometers of track, serves about ten million fewer people, and competes with eight other companies. Japan’s railway system turns a large operating profit and receives far less public subsidy than European and American railways.

In most developed countries, the railways have struggled since the rise of the automobile in the 1950s. From this point on, North America saw the near-total replacement of passenger trains with cars and planes. In Europe, it meant vast government financial support to keep the lines open.

Japan’s different trajectory is often attributed to culture: the Japanese are conformists who are content to take public transport, unlike freedom-loving Americans who prefer to drive everywhere. Europeans are somewhere in between. Culture is also used to explain the incredible punctuality of Japanese railways.

These cultural explanations are wrong. The Japanese love cars, but they take trains because they have the best railway system in the world. And their system excels because of good public policy: business structure, land use rules, driving rules, superior models for privatization, and sound regulation have given Japan its outstanding railways.

This is good news for friends of rail. Culture is built over centuries, and replicating it is hard. But successful public policies can be emulated by one good government. Much about Japan’s railway system could be replicable around the world.

Japan’s railway companies

Today, the most striking institutional feature of Japanese rail is that it is privately owned by a throng of competing companies.

The railway arrived in Japan in 1872, during the Meiji Restoration, which opened the country up to foreign trade, ideas, and technologies. Like most Western countries, Japan nationalized its railways in the early twentieth century, creating what became known as Japanese National Railways (JNR). But it did not nationalize all of the lines, focusing only on mainline railways of national importance, and new private railways were still permitted.

Between 1907 and World War II, Japan saw a boom in new private electric railways, coinciding with rapid urbanization. Technologically, most of these private railways were similar to the famous interurbans in the United States: they were basically electric trams, but running between cities as well as within them. The American network eventually withered, and almost nothing of it survives today. In Japan, however, the network consolidated, and the light tramlines gradually evolved into heavy-rail intercity connections.

These companies are today known as ‘legacy private railways’ on account of their having been private since their inception. There are eight legacy private railways in the Tokyo metropolitan area, five in the Osaka–Kobe–Kyoto megalopolis, two in Nagoya, and one in the fourth city of Fukuoka. There are also dozens of smaller ones elsewhere. In the three largest urban areas, these operators account for nearly half of railway track and stations, as well as a plurality of ridership. The largest, Kintetsu, not only operates urban services, but a whole intercity network stretching from Osaka to Nagoya.

These companies often compete head-to-head. At its most extreme, three separate commuter lines compete for the traffic between Osaka and the port city of Kobe, running in parallel, sometimes fewer than 500 meters apart.

Meanwhile, the nationalized railways were managed by JNR. In the postwar era, JNR was responsible for building the famous Shinkansen system, as well as running commuter and long-distance lines throughout Japan. But in 1988, it was largely privatized, broken into six regional monopolies for passenger services together with a single national freight operator. These are collectively known as the Japan Railways Group (JR).

This means that Japan has ended up with six railway companies that trace their descent to the nationalized railways, the sixteen big legacy companies that have always been private, and a host of minor legacy railways, as well as numerous underground metros (some private, some municipally owned), monorails, and tram systems. This institutional diversity is striking enough. But equally striking is the consistent business model that has evolved amidst this pluralism: the railway that builds a city.

Railway-led urbanism

If I take a train to go for a solitary walk in the countryside, the railway company can capture some of the value it creates by charging me for the journey, just as other companies capture the value of the goods and services they provide by charging for them. However, if I take a train to visit family, clients, a theater, or a shop, an important difference appears. The railway can capture the value it creates for me by charging me a fare, but it cannot capture the value it creates for those at my destination. As transport infrastructure creates benefits that produce no revenue for providers, free markets rarely build enough of it.

Japan has partly solved this problem by enabling railway companies to do a great deal beside running railways. Take the example of the Tokyu corporation, one of the legacy private railways in southern Tokyo. You can not only travel on its trains, but also ride a Tokyu bus, live in a Tokyu-built house, work in a Tokyu office complex, see a doctor in a Tokyu hospital, buy groceries in a Tokyu supermarket, spend an afternoon at a Tokyu museum-theater-cinema complex, take your children to their amusement park, and even die in a Tokyu retirement home. The positive spillover effects of the railway on these things are captured by Tokyu because it owns them. The president of Tokyu has said:

I think that though we are a railway company, we consider ourselves a city-shaping company. In Europe for instance, railway companies simply connect cities through their terminals. That is a pretty normal way of operating in this industry, whereas what we do is completely different: we create cities and then, as a utility facility, we add the stations and the railways to connect them one with another.

This model was pioneered in the 1950s by what became Hankyu Railways. Hankyu’s network connects central Osaka to its northern suburbs, as well as Kyoto and Kobe. Its innovative founder Kobayashi Ichizo first built suburban housing, then a department store at the terminal station; he then created a hot spring resort, a zoo, and his own distinctive brand of all-women musical theater, the Takarazuka Revue. He also began to run bus services to and from his stations. Other companies emulated Hankyu’s example: Tokyo Disneyland is a collaboration between Disney and the Keisei Railway, while Hanshin in Osaka owns the Hanshin Tigers baseball team.

Core rail operations are profitable for every Japanese private railway company, but they usually only account for a plurality or a small majority of revenue. The rest is contributed by their portfolio of side businesses. There is a natural financial synergy between the reliable but unremarkable cash flow of train fares and the profitable but riskier real estate and commercial side of the business. Railway companies’ side businesses also attract people to live and work on their rail corridor, reinforcing the customer base for the railway services themselves.

This virtuous circle is enabled by transit-oriented development. Japan’s liberal land use regulation makes it straightforward to build new neighborhoods next to railway lines, giving commuters easy access to city centers. It also enables the densification of these centers, which means that commuters have more places they want to go.

Railways cost a lot to build, but once they are built, they can move enormous numbers of people, far more than a road of similar size. This means that they work best in cities with a high density of people, jobs, and other activities. In 2019, New York City was the only American city where rail had a higher modal share than cars, in part because Manhattan has 2.5 million jobs, two million residents, and 50 million tourist visits crammed into 59 square kilometers.

This does not mean that rail-oriented cities must be structured like Chinese cities: islands of high-rise apartments connected by metros and separated by motorways. Japanese cities have the lowest residential density in Asia, and a plurality of the Japanese live in houses, usually detached ones. The urban area of Tokyo, the densest Japanese city, has a weighted population density less than that of many European cities, including Paris, Madrid, or Athens. Japanese cities have vast low-rise, predominantly residential suburbs, built at densities that might be higher than what is typical in the United States, but that would be quite normal in Northern Europe.

What makes Japan’s cities particularly suited to rail is thus not their residential districts, but their huge and hyperdense centers. These really are special: the cores of Tokyo or Osaka are unlike anything that exists in Europe or North America. Many of their features are famous worldwide: the vertical street zakkyo buildings, underground streets, shopping streets under rail tracks, covered arcades, elevated station squares, and vertical cities. Getting millions of commuters and shoppers into these downtowns is where rail excels because its extreme spatial efficiency means that infrastructure with a relatively modest footprint can transport vast numbers of people into a small area.

None of this emerged from a coherent masterplan of transit-oriented development like Copenhagen’s Finger Plan or Curitiba’s Trinary System. Postwar Japanese opinion was committed to decentralization both to rural peripheries and to the suburbs through greenbelts, motorways, and new towns.

Instead, this variety and adaptability around railways is possible because of the way Japanese urban planning works. Since 1919, Japan has had a standardized national zoning system, but it is much more liberal than development control systems in Western countries. The Japanese authorities did not intend or even desire dense urban centers, but they did not prevent them, rather like nineteenth-century governments in the West.

This liberal zoning system is reinforced by private access to city planning powers. Thirty percent of Japan’s urban land has been subject to land readjustment, where agreement among two thirds of residents and landowners in an area is enough to allow its replanning, including compulsorily taking and demolishing land for amenities and infrastructure. Initially land readjustment was used only to assemble rural land for urbanization, but over time it was increasingly used to redevelop already urbanized areas, and new variants were created to build the skyscrapers that surround the major stations of central Tokyo.

The history of the private railway companies could be written as a story of land readjustment projects: the initial building of the lines in the interwar years proceeded through one land readjustment project after another. Postwar improvements such as double-tracking, platform lengthening, and constant redevelopment of stations and their immediate thresholds were only possible because the railways could secure land takings cooperatively with local businesses and landowners.

Perhaps the greatest example of this phenomenon involved Tokyu. In 1953 the company decided to build the Den’en Toshi Line, or Garden City Line, to serve a rural area southwest of Tokyo. This would be enabled by a series of land readjustment projects collectively among the largest in Japanese history.

Over 30 years, 3,100 hectares were covered, of which only 36 percent was devoted to residential and commercial development, with 20 percent for forest and parks, 17 percent for roads, and much of the rest for watercourses. The population of the land readjustment zone would rise from 42,000 in 1954 to over 500,000 in 2003.

By connecting the affluent southwestern suburbs to Tokyu’s main real estate hub next to Shibuya station, now the second busiest in the world, the Den’en Toshi Line allowed Tokyu to become the largest private railway by revenue and ridership. The Japanese government and academics generally consider the Den’en Toshi Line to be the best corridor of transit-oriented development in Japan.

But the railway-as-city-builder model is not the only reason Japanese railways have been able to thrive. European countries usually prohibited railways from running real estate side businesses, but in the United States and Canada the practice was extremely widespread in the nineteenth and early twentieth centuries, and many famous railway suburbs were developed this way. Despite this, passenger rail in these countries collapsed in the mid-twentieth century. Part of the difference was that Japan did not extend the same implicit subsidies to cars as Western governments did.

Pricing driving

The land of Toyota, Nissan, and Honda is not an anti-car nirvana. In fact, Japan has excellent motorways, and across the country as a whole a small majority of journeys are made by car. But Japan is a place where cars and car-oriented lifestyles compete on a level playing field.

Japan is one of the only countries to have privatized parking. In Europe and North America, vast quantities of parking space is socialized: municipalities own the streets and allow people to park on them at low or zero cost. Initially with the intention of encouraging the provision of more parking spaces, Japan made it illegal to park on public roads or pavements without special permission. Before someone buys a car, they must prove that they have a reserved night-time space on private land, either owned or leased.

Since parking on public land is banned, municipalities are not worried about overspill parking from developments with inadequate private parking. They therefore have no reason to impose parking minimums on developments: the market is left to decide whether parking is the most valuable use of private land. Where land is abundant, as in rural areas, suburbs, or small towns, private parking is plentiful. But in city centers, it is outcompeted by other land uses. According to the late Donald Shoup, central Tokyo has 23 parking spaces per hectare and 0.04 parking spaces per job, compared with 263 and 0.52 for Los Angeles. Even Manhattan, the densest urban area in North America with the lowest levels of car ownership, has about 60 parking spaces per hectare.

Japanese roads are expected to be self-financing. Motorways are run by self-contained public cooperatives, very similar to the statutory authorities that ran English roads and canals between 1660 and the late 1800s, and funded by tolls on their users. Vehicle registration taxes, which are allocated to localities for road construction and maintenance, are worth three percent of the Japanese government budget.

These measures, adopted in the 1950s, were not intended to suppress car use – the point was to fund a massive road expansion – but they have forced private vehicles to internalize many of their hidden costs. In the Tokyo urban area, the average household spends 71,000 yen ($450) each year on public transport fares and 210,000 yen ($1,350) on car purchase and maintenance costs.

But the private car was not the only competitor faced by the private railways. For eight decades in the twentieth century, they also had to face the juggernaut of Japanese National Railways. Its privatization in 1988 removed the final obstacle to creating the world’s best railway system.

Privatization

Railway privatization in Britain, New Zealand, Argentina, and Sweden has had a mixed reception, and all of those countries, apart from Sweden, have taken steps to reverse it. In Japan, it has been so successful that the government subsequently privatized the metro systems in Tokyo and Osaka.

In the postwar period, JNR enjoyed real successes. It built the revolutionary Shinkansen, the first high-speed railway in the world. It also aggressively electrified and double-tracked major trunk lines, quadruple-tracked lines into and out of major cities, and added city-center loops and freight bypasses. But these achievements were overshadowed by two problems.

The first was politics. Many countries adapted to the rise of the car by closing the least profitable parts of their passenger rail network, like the consolidation of American freight rail into the Class I operators or the Beeching Axe in Britain. In Japan, however, the ruling Liberal Democratic Party drew its support from rural constituencies, whose support it retained with pork-barrel politics. Its ‘rail tribe’ group, led by rural MPs, prevented JNR from adapting itself to mass motorization.

JNR therefore did not amputate gangrenous rural and freight services that imposed heavy costs with few benefits. Worse, it continued to build new loss-making rural railway lines, known in Japanese as Gaden-intetsu, or railways pulled into the rice field.

The second problem was organized labor. In general, Japanese trade unions are known for their moderation and responsibility, a generalisation that also held true for the unions at the legacy private railways. The JNR unions, however, became highly militant, secure in the knowledge that their nationalized employers could never go bankrupt. Their largest series of strikes in 1973 provoked riots from commuters.

The railway unions imposed overstaffing on revenue-generating urban services, at a time when both international and private domestic operators were reducing staffing requirements against a backdrop of higher wages and the growing automation of signaling and ticketing. As a result, 78 percent of JNR’s costs were related to labor, compared to 40 percent for other Japanese railways. The average worker at a private railway was 121 percent more productive than their JNR counterpart.

By the early 1980s, only seven out of 200 JNR lines made a profit. Successive governments deferred serious reform, running up debt, cutting down investments in new urban lines, raising ticket prices to twice those of comparable private railways, and increasing subsidies – which rose until annual subsidies equaled the total cost of the Shinkansen.

In 1982, Prime Minister Yasuhiro Nakasone started to privatize the railways. Unlike other countries, Japan simply returned to the traditional private railway model of the nineteenth and early twentieth centuries: tracks, trains, stations, and yards were owned by vertically integrated regional conglomerates.

There are substantial advantages to vertical integration. Railways are a closed system that has to be planned as a single unit. Changing the timetable at station A can affect the timetable at station Z; buying new trains that can travel faster might require changes to the infrastructure so they can reach their top speed, which in turn requires rewriting the timetables. This becomes especially complicated if different services share tracks. To prevent delays from propagating from one service to another, the timetable needs to be carefully designed to make best use of the available infrastructure.

The starkest effect of privatization was a massive and immediate increase in labor productivity and profitability relative to the legacy private railways. In fact, this began before privatization: its mere threat strengthened the government’s hand when bargaining with the unions and forced JNR to begin closing rural lines.

Privatization saw a general trend of productivity improvements, following a big one-time improvement between 1982 and 1990, when the workforce was cut by more than half, 83 loss-making lines were removed, and JNR’s debts were transferred to a holding company.

The second great advantage of privatization was to allow the JR companies to emulate the railway-as-city-builder model of the legacy private railways: for instance, JR East owns two shopping center brands, a ski resort, a coffee chain, and even a vending machine drink company. The JR companies have not ignored their rail business: they have continued to build new high-speed lines and urban tunnels, upgrade stations, and implement a host of other improvements such as the introduction in the 1990s of smart cards that allow passengers to pay their fare with a tap.

Regulation

This does not mean that the Japanese railway industry is a pure creature of free enterprise. No railway system ever has been. The Japanese system has found an equilibrium that makes rail policy explicit and limited. Leaving aside railway safety and business regulation, there are two main policy levers: fare maximums and capital expansion subsidies.

Price controls are often cited as a classic example of misguided government intervention, whether through rent controls, caps on the price of gasoline, wage freezes, or minimum agricultural prices. Tokyo’s infamously crammed trains are a symptom of underpriced rush hour traffic.

Railways have market power because the substitutes for railway trips – coaches, cars and planes – are quite a different product. This monopolistic position has historically meant trouble: monopoly systems, whether private or public, have a tendency to abuse their position to charge higher prices and run bad services. For this reason, the private monopolies that were common in the Western world before World War I often had price controls imposed on them. For example, most of the American streetcar networks were operated as long-term, price-controlled franchises granted by the city.

Price maximums, if set too low, could have ruined Japan’s railways. This is exactly what happened to many Western transit services after the First World War. But the postwar Japanese practice has capped fares generously. The system is explicitly designed to maintain profitability per rider, which in turn incentivizes the companies to maximize ridership. That buys political legitimacy for the privatized system, which is necessary for the continued provision of capital expansion subsidies. Indeed, during the long deflation era between 1992 and 2022, it was common for operators to charge below the maximum, and the real value of railway fares continued to rise. Fare maximums are set on the basis of the average cost structures of all railway operators in a region, so companies with below-average costs like Tokyu would often charge below the cap to maintain a competitive edge, prevent public backlash, and maximize traffic to their side-businesses.

Other than the fare maximums, the railways are free to make their own decisions about timetables, service patterns and day-to-day operations, a highly specialized and technical task which requires deep expertise. This contrasts with the government meddling with, say, Amtrak’s routes.

Carefully designed public subsidies also play a useful role. Although Japanese railways do not receive subsidies for day-to-day operations, they do receive government loans and grants for capital investments. These are typically tied to public priorities, such as disability access or earthquake-proofing, or to projects that have large spillovers that the railway company would be unable to internalize, like removing level crossings, or elevating at-grade railways or trams in order to reduce road congestion and accident risk. Generally, the local prefectural government will match the contribution of the national government. Larger new build projects are subject to lease back or debt-payment conditions that fare revenue is expected to pay back.

Subscribe now

The recipe for successful railways

Railway companies invested heavily in real estate businesses, often funding lines through selling land for housing around new stations. Liberal spatial policy meant that such development happened easily, even as it enabled dense development in urban cores where radial rail lines converged. Rail companies were generally vertically integrated regional monopolies, owning the land, track, and rolling stock, setting their own timetables, and employing their staff. The state imposed controls to stop them exploiting their monopoly position, but it did so cautiously, allowing them to make sufficient profit that incentives to invest were preserved. Capital subsidies were targeted at providing specific public goods that normal commercial operations overlooked.

The above paragraph could be written by a historian of the future about contemporary Japan. But every word in it could also be written by a historian today about the United States in the nineteenth century – usually seen as the epitome of capitalist individualism. This striking fact contradicts the idea that America’s supposed individualism foreordains it to be the land of the car, or that Japan’s supposed communitarianism foreordained it to be the land of rail.

It also puts pressure on the idea that the demise of rail is the inevitable consequence of cars. All countries saw some shift to cars in the twentieth century, and all rail industries had to respond to that. But public policy had an enormous effect on how successfully they did so. The rise of zoning restrictions on density, excessive price controls, nationalization, and vertically disintegrated privatization have hampered Western rail in remaining competitive against cars since the 1920s. By maintaining and restoring the institutions that built the first railway systems in the nineteenth century, the Japanese have created the mightiest railway system of the twenty-first.

Matthew Bornholt is an urban planner and transport researcher. You can follow him on Bluesky.

Benedict Springbett is a writer and Bar student. You can follow him on Twitter.

This is the second article of Issue 23, which print subscribers will start receiving this week. Not yet a subscriber? You can sign up for the magazine here.

The convenience of instant coffee masks a surprisingly difficult problem. Coffee’s appeal lies in the hundreds of volatile compounds that create its flavor and aroma, exactly the substances most likely to disappear during processing. Creating instant coffee required developing techniques to extract the soluble molecules in coffee from the insoluble plant matter without destroying the fragile compounds that make coffee worth drinking.

The first attempt at the drink was, by all accounts, terrible. In 1771, over two centuries after coffee reached Europe, Londoner John Dring filed a patent for a ‘coffee compound’. Dring’s method involved mixing ground coffee with butter and tallow, then heating the mixture on an iron plate until it thickened into a paste that could be shaped into cakes. These cakes were then dissolved in hot water to make coffee. The purpose of the animal fats isn’t entirely clear. They may have been intended to extract and carry soluble compounds from the coffee grounds or to preserve the ground coffee from oxidation. Whatever Dring’s aim, the method wasn’t commercially viable because the fats went rancid, causing the cakes to spoil quickly.

During the mid-1800s, several firms produced instant coffees as thick liquid concentrates that could be reconstituted with water. In 1840, the Scottish company T & H Smith developed a ‘coffee essence’ by brewing coffee and reducing it to around a quarter of its original volume. This thick liquid was mixed with chicory extract and burnt sugar syrup, creating a molasses-like concentrate. One or two teaspoons mixed with boiling water made a drink, though it tasted more like coffee flavored molasses than proper coffee.

Another attempt came during the American Civil War. In 1832, President Andrew Jackson had replaced soldiers’ daily spirit rations with coffee beans and sugar. This created a heavy logistical burden for the army, with a 20-day supply for 100,000 troops weighing 250 tons, all needing transport by horse-drawn wagon. Roasting, grinding, and brewing coffee in the field was also time-consuming for soldiers.

In 1861, the Union Army began investigating instant coffee as a solution. They procured a coffee concentrate from the firm HA Tilden & Co, consisting of a mixture of thickened coffee and sweetened, condensed milk. This halved the weight and size of the coffee, but was unpopular with soldiers, who compared its consistency to axle grease.

These essences were made by boiling down brewed coffee to concentrate it. This damages the flavor, producing a bitter, unpleasant drink, hence the old saying that ‘coffee boiled is coffee spoiled’. That’s why they were syrups. Boiling away all the water to create a dry powder would have destroyed whatever coffee flavor remained. To make a viable instant coffee powder, producers needed a way to remove the water from brewed coffee without boiling it.

A spice merchant’s solution

The first genuine instant coffee powder emerged in 1889, created by David Strang, a spice merchant in Invercargill, New Zealand. He developed a ‘Dry Hot-Air’ method that removed water from coffee by blowing heated air over it, likely using a spice dryer he’d patented a few years earlier. While hot air dehydration had been used in France since 1795 to dehydrate foods like pasta for commercial sale, Strang was the first to apply it to coffee.

The method works by warming the air around the coffee rather than the coffee itself. Turning water into vapor requires energy, which evaporating water draws from the heated air passing over it. Because this energy comes from the air rather than the liquid, the coffee surface actually gets cooler as it evaporates. This keeps the liquid below boiling temperature even as it dries.

As a dry powder, it was lighter and more shelf-stable than previous attempts at instant coffee, making it more practical for shipping and storage. The product achieved modest commercial success in New Zealand, being advertised as ‘Far superior to any so-called coffee essence’. But the taste was still far from ideal, coffee expert Arjun Haszard notes: ‘This process, given what we know about what happens to coffee with heat and air would have undoubtedly resulted in heat damaged, oxidized coffee. Portable, yes, but also most likely horrible’.

George Washington goes to war

The first instant coffee to achieve widespread commercial success came in 1909 when Belgian-British inventor George Constant Louis Washington launched Red E Coffee. Washington kept his method a trade secret, so the exact process remains unknown, but the result wasn’t clearly superior to previous attempts, with the taste described as ‘disagreeable’. Washington’s success seems to have come mainly from being the first to build an industrial-scale production facility, located at Brooklyn’s Bush Terminal industrial complex.

The coffee was fairly popular in its first few years of production, but demand surged when World War I began. The military procured the entire available production, which peaked at 37,000 pounds (16.7 metric tons) per day. Despite the poor taste, the sheer convenience of instant coffee provided a significant morale boost to the troops. One soldier’s letter home captures this sentiment:

I am very happy despite the rats, the rain, the mud, the draughts, the roar of the cannon and the scream of shells. It takes only a minute to light my little oil heater and make some George Washington Coffee . . . Every night I offer up a special petition to the health and well-being of Mr. Washington.

By the end of the war, the American army was being called the best-fed on Earth. Instant coffee rations, no doubt, were an enviable luxury.

Solving the sticky problem

In 1929, the Wall Street Crash devastated Brazil’s economy. Brazil relied heavily on coffee exports, which made up half of its total exports, and the US was its biggest customer. Coffee prices collapsed 90 percent within a year, contributing to the 1930 revolution that overthrew the government. To stabilize prices, the Brazilian government burned 10.3 billion pounds (4.6 million metric tons) of coffee over multiple years. This was the equivalent of three years’ worth of global coffee production. A 1937 Time article described the scene: ‘huge grey-green piles of coffee beans smouldering slowly away under great smoke plumes, barges lumbering out to sea to dump coffee overboard, workmen mixing coffee and tar into briquets for building’.

During the crisis, Banque Française et Italienne pour l’Amérique du Sud, the French and Italian Bank for South America, found itself with a huge surplus of coffee in its warehouses. To find a use for the excess beans, they approached Nestlé chairman Louis Dapples, a former employee of the bank, with a proposal to develop a better instant coffee product.

Swiss chemist Max Morgenthaler was put in charge of the project in 1932, but with no signs of success, Nestlé cut funding for it in 1935. Undeterred, Morgenthaler purchased his own coffee beans and continued his research from home, occasionally using equipment from the factory laboratory during quiet periods. In April 1937, he achieved a breakthrough and presented samples to Nestlé’s executive board. They were well received. One attendee exclaimed, ‘Mother Nestlé has produced a beautiful baby!’

Morgenthaler’s process involved passing hot water through multiple columns of ground coffee to create a coffee extract, which was then spray dried. Spray drying is a technique invented by chemist Samuel Percy in 1872 for converting liquid into powder. It works by spraying liquid as a fine mist into a heated chamber, where hot air rapidly evaporates the water, turning the droplets into powder.

Spray drying had been successfully used to make milk powder since the early 1900s, but applying it to coffee presented unique challenges. Coffee’s natural sugars and acids have low molecular weights, meaning they soften and become sticky at relatively low temperatures. During spray drying, this causes the coffee to clump together into a paste rather than forming a free-flowing powder. Morgenthaler’s solution was to mix the coffee with roughly equal parts carbohydrates like maltodextrin or glucose before drying. These larger carbohydrate molecules remain solid at higher temperatures, raising the point at which the mixture becomes sticky and allowing it to dry into proper powder particles.

Despite using much higher air temperatures than previous methods (typically 150–250 degrees centigrade), spray drying actually causes less heat damage to the coffee as the droplets spend only seconds in the heated chamber, and the evaporating water keeps them relatively cool.

The product was launched in 1938 as Nescafé. It was an instant hit, and a year’s worth of stock sold out in two months. World War II boosted sales massively. In 1942, demand from the US military was so great that the government classified it as a ‘commodity vital to the war effort’, with the entire output of Nestlé’s US plant dedicated to supplying the armed forces.

From ice to vapor

While Morgenthaler solved the technical problem of spray drying coffee, the added carbohydrates diluted the coffee flavor significantly. Nestlé resolved this in 1952 by discovering they could extract natural carbohydrates from the coffee itself. The method worked by first passing water through ground coffee at very high temperatures (up to 175 degrees centigrade) under pressure, breaking down cell walls and releasing carbohydrates called polysaccharides that don’t dissolve at normal brewing temperatures. The liquid then flows through additional columns at around 100 degrees centigrade to extract flavor compounds, as in the original method. The result was 100 percent pure coffee powder.

Spray drying kept heat exposure brief, but manufacturers wondered if they could avoid high temperatures altogether using a process called freeze drying. The method involves freezing materials, then placing them under low pressure, turning the ice into vapor without passing through a liquid phase. This happens because at very low pressures, water can exist only as solid ice or as vapor, not as liquid, so gentle warming turns the ice straight into vapor, a process called sublimation.

Various cultures had practiced forms of freeze drying for centuries – the Inca freeze dried potatoes in the high Andes in the thirteenth century – but industrial interest grew during World War II when trying to find a way to preserve blood plasma and penicillin. Without reliable refrigeration, these supplies often spoiled before reaching patients. Freeze drying removed the water content, allowing them to be stored at room temperature and quickly reconstituted when needed.

In 1963, Maxwell House, a coffee brand owned by General Foods (and now part of Kraft Heinz) released the first freeze-dried instant coffee. Nestlé followed in 1965 with Nescafé Gold. The process involves freezing coffee extract to between minus 40 and minus 45 degrees centigrade, and then breaking the frozen slabs into granules and sieving them to size. These granules are placed in vacuum chambers where controlled heat causes the ice to sublimate over several hours, leaving dry coffee powder.

Freeze drying retains flavor much better than spray drying. One study found that freeze drying maintained 77 percent of volatile compounds compared with 57 percent for spray drying.

Another advantage is texture. Freeze drying produces coarse, porous granules that dissolve quickly and completely in hot water. Spray drying, by contrast, produces fine powder particles that tend to float on the surface rather than dissolve. To address this, manufacturers now often add an agglomeration step where spray-dried powder is exposed to steam as it tumbles in the air. The moisture causes particles to stick together and form larger granules.

Freeze drying has drawbacks. The process takes 8–16 hours and requires expensive vacuum equipment and condensers. Freeze-dried coffee typically retails for roughly twice as much as spray dried coffee, which is why spray drying remains the dominant method.

Premium instant

Instant coffee has spent most of its history as the cheapest, quickest, and most portable coffee, but with a reputation for low quality when it comes to the flavors that coffee lovers seek out. That has begun to change: a market for premium instant coffee has opened up over the past two decades. Today, specialty roasters like Verve and Supreme offer freeze-dried versions of their coffees, often selling for around $2.50 per cup, 35 times the price of standard instant.

Making this possible required technical breakthroughs. One issue was aroma loss. While freeze drying helped preserve more volatile compounds during drying, delicate aromatics could still be lost during earlier stages like roasting, grinding, and extraction. Retaining more of these compounds required improved aroma recovery methods that capture volatiles early in the process, store them separately, and add them back after drying. Primitive forms of aroma recovery had existed since the early twentieth century, but advances since the arrival of freeze drying gave manufacturers better tools to preserve the subtle characteristics that distinguish specialty beans.

Flavourtech’s spinning cone column, originally developed for dealcoholizing wine, has become a popular aroma recovery technology in premium instant coffee production. The device consists of a vertical cylinder containing alternating fixed and rotating cones stacked like ice cream cones. A slurry of ground coffee and cold water poured in at the top cascades down, with each rotating cone flinging the liquid outward into a film roughly a millimeter thick. Low-temperature steam (40–50 degrees centigrade) rises through the column, collecting volatile aromatics from the coffee before cooling and condensing into concentrated aroma extract. The process is quick: coffee spends only 25 seconds in the machine.

Even with gentler methods, flavor compounds are inevitably lost during the drying process. Flash-frozen coffee avoids this by skipping dehydration entirely. Flash freezing is an existing technique, but was adapted to coffee by Massachusetts-based company Cometeer, founded in 2016. It involves brewing coffee at ten times normal strength, then exposing it to cryogenic temperatures using liquid nitrogen. At this extreme cold, the coffee freezes almost instantly, fast enough that ice crystals remain small and don’t damage the coffee’s structure or degrade volatile compounds. The frozen concentrate is then sealed in aluminum capsules.

Flash freezing isn’t cheap. It requires expensive cryogenic equipment and frozen distribution infrastructure, because products must be packed in dry ice and delivered quickly to avoid thawing on the way to the end consumer’s freezer. As a result, Cometeer’s capsules start at around $2 per cup, with prices reaching $7.50 for the most premium beans.

Texture and mouthfeel presented another opportunity for improvement. Fresh-brewed coffee contains oils that coat the tongue and fine particles that create body. The water extraction process used to make instant coffee leaves these behind. Because they don’t dissolve in water, they can’t be captured in the soluble powder. Without them, instant coffee tastes thinner.

Microgrounds offered a way to address this. In September 2009, Starbucks launched VIA Ready Brew, a product that had been in development for nearly 20 years under Don Valencia, the founder of Starbucks’ R&D division. The innovation combined soluble instant coffee with roasted coffee particles small enough to suspend in water rather than fully dissolve. Competitors quickly developed their own microground offerings, with Jacobs Douwe Egberts (now JDE Peet’s) launching Millicano in 2011 and Nestlé launching Nescafé Azera in 2012. However, this hasn’t been universally adopted in premium instant coffee. One issue is that the microgrounds leave a layer of sediment at the bottom of the cup.

Technology alone wasn’t enough to create a premium instant market. The economics of production also had to change. Instant coffee production requires multi-million dollar capital investments in extraction batteries, concentration equipment, and drying facilities. Historically, only manufacturers running at massive scale could justify these costs, leaving specialty roasters with no realistic path into the category.

This changed in 2016 when Nate Kaiser founded Swift Cup Coffee in Lancaster, Pennsylvania, pioneering what you might call instant coffee as a service. Today, a vibrant market of contract processors exists to serve specialty roasters. These processors brew roasters’ beans to precise extraction standards, freeze dry in small batches, and package the finished product under the roaster’s own brand. This converts lumpy fixed costs into variable costs, letting roasters test the market without major investment.

While instant may never be the coffee connoisseur’s preferred drink, decades of innovation have earned it a role in millions of people’s lives. From troops in the field to rushed mornings and camping trips, it offers a practical solution when time or equipment are scarce.

Benjamin Stubbing is an economic analyst. Oscar Sykes is a software engineer. You can follow him on Twitter.