You Get What You Pay For: How Incentives Rule the World.

Part one of a three-part series on incentives, capital allocation, and the machine being built to fix both. This part: how the West wired itself to reward exactly the things it says it doesn’t want.

In 1902, the French colonial government in Hanoi had a rat problem. The city’s shiny new sewer system had become a subterranean rat paradise, and with bubonic plague drifting through Asia, the administration decided to act. They did what governments do: they created an incentive. One cent per dead rat. To keep things hygienic, you didn’t have to hand in the whole rat. Just the tail.

The scheme worked magnificently, if you were a rat.

Within weeks, the bounty office was drowning in tails. Nearly 8,000 a day by late April. Over 15,000 a day by the end of May. North of 20,000 a day by June. And yet, strangely, Hanoi did not seem to be running out of rats. Then health inspectors began noticing something odd in the streets: live rats, perfectly healthy, scurrying about with no tails. The rat catchers had worked out that a dead rat pays once, but a live rat with its tail snipped off goes home and makes more rats, and every one of those rats has a tail. Out in the countryside, enterprising Vietnamese had gone one better and set up rat farms. Breeding the things. For the bounty.

The French cancelled the programme in disgust. The following year, plague arrived in Hanoi anyway.

The story is usually told as a joke about colonial incompetence. It is actually a controlled experiment in the single most powerful force in economics. The rat catchers of Hanoi were not stupid or lazy. They were brilliant. They met the incentive in front of them with creativity, entrepreneurship and ruthless efficiency. The problem was never the people. The problem was that the incentive paid for tails when what the government wanted was fewer rats.

Hold that thought, because the modern West is one giant bounty office overflowing with rat tails, paying out in a hundred currencies: hotel nights, sickness notes, renovation credits, unprosecuted theft. And there is now, for the first time, a working technology that does the opposite: a machine that pays only for the thing you actually want, measures it relentlessly, and pulls capital away from failure in real time, with no committee anywhere in the loop. We will meet it properly later in this series. First, an honest tour of the wreckage.

Charlie Munger spent seventy years repeating the same sentence: “Show me the incentive and I will show you the outcome.” He also said that he had been in the top five per cent of his age cohort in understanding incentives all his life, and had still underestimated them every single year. Steven Levitt and Stephen Dubner built Freakonomics, the best-selling economics book of the century, on the same foundation: economics is, at root, the study of incentives. Their tour of incentives misfiring ran through classrooms, dojos and estate agencies. This one runs through national economies. If the sharpest capital allocator of the twentieth century couldn’t stop underestimating incentives, and the best-selling economists of the twenty-first filled two volumes with clever people gaming them, what chance does a parliamentary subcommittee have?

Not much, as it turns out.

Pay people to arrive, and they will arrive

Start with the example currently lighting up podcast land: illegal migration. Strip away the politics and look at it purely as mechanism design. If you are a rational person in a poor or dangerous country, and a rich country’s policy stack says that arrival triggers accommodation, healthcare, a weekly allowance and a years-long legal process during which removal is unlikely, then that policy stack is a bounty. Not morally. Mechanically. It pays out on arrival. And bounties get harvested.

The numbers are what you’d expect. In 2023-24 the UK spent £4.7 billion on asylum support, £3 billion of it on hotels, roughly £8 million a day. The ten-year asylum accommodation contracts, originally costed at £4.5 billion, have been re-costed to £15.3 billion. A hotel place runs about £170 per person per day against £27 in dispersal accommodation, and in 2024/25 hotels held about a third of supported asylum seekers while eating three quarters of the accommodation budget. Meanwhile some 41,500 people crossed the Channel in small boats in 2025, the second-highest year on record. Whatever you think the compassionate policy is, nobody sat down and designed this outcome. It emerged, the way tailless rats emerged, from well-meaning rules meeting rational people.

New York ran the same experiment at American scale. A 1981 consent decree gave the city a legal right-to-shelter obligation, which in 2022-25 collided with two hundred thousand new arrivals. Result: no-bid emergency hotel contracts at around $373 a night per household, the Roosevelt Hotel reborn as a migrant intake centre, and roughly $8 billion of spending in three fiscal years. When arrivals collapsed after 2024’s border policy changes, the shelter census fell from 69,000 to under 37,000 and the Roosevelt closed. The tap was the incentive. Turn it off and the queue shortens. Economists will argue forever about how much of the pull was the shelter guarantee versus the border itself, but no serious person now argues the incentive was a bystander.

Sweden completes the set. In 2015 it took 163,000 asylum applicants, the most per capita in the OECD, on an explicit “open your hearts” policy. Ten years later the same country recorded net emigration for the first time in half a century and now offers voluntary repatriation grants of 350,000 kronor per adult, about $34,000, to leave. Sweden is running the bounty in reverse. Denmark next door skipped the whole cycle by making itself deliberately unattractive, and granted 864 asylum approvals in 2024, near a forty-year low. Same continent, same treaties, same decade. The only variable that moved was the incentive.

Pay people to be ill, and they will be ill

The NHS is free at the point of use, and the British are rightly attached to the principle. But a price of zero is still a price, and it does what all prices do: it coordinates behaviour. Make a thing free and demand is rationed by queue instead, and the queue becomes the product. NHS England’s waiting list peaked at 7.77 million cases in September 2023, against 4.57 million before the pandemic. It has since been ground down to just over 7 million, and hitting “65 per cent of patients seen within 18 weeks” was announced as a triumph. Eighteen weeks. As a target. And because a thing that costs nothing to book costs nothing to skip, more than a million GP appointments a month are simply not attended.

The really spectacular UK example is what happened when the state started paying materially more for sickness than for job-seeking. Health-related benefits are worth more than unemployment benefits, carry no work-search conditions, and in the case of PIP aren’t even means-tested. After 2020, new PIP awards went from about 13,000 a month to about 34,000 a month. The Institute for Fiscal Studies counted 4.2 million working-age people on health-related benefits in 2023-24, up from 3.2 million in 2019, with spending headed from £48 billion toward £64 billion by decade’s end. Economic inactivity due to long-term sickness sits around 2.8 million people, up from 2 million in 2019. Did the British population’s underlying health collapse by forty per cent in five years? The IFS itself notes that health deterioration alone cannot explain the surge. And when the government tried to trim the incentive in 2025, its own backbenchers gutted the reform in a fortnight. That is the other thing about perverse incentives: they build constituencies. The bounty defends itself.

America runs the mirror-image experiment and gets the mirror-image pathology. The US pays for healthcare by the procedure, so it gets procedures. Five point three trillion dollars of them in 2024, eighteen per cent of GDP, $15,474 for every man, woman and child, in exchange for a life expectancy about four years shorter than comparable countries. A JAMA study put the waste at up to $935 billion a year. My favourite micro-example: until 2011, Medicare paid dialysis clinics cost-plus for an anaemia drug called EPO, and clinics duly overdosed their patients with it, more than $2 billion a year, cardiovascular risk evidence notwithstanding. In 2011 the payment was flattened into a bundle, and usage dropped by a third almost overnight. The patients hadn’t changed. The science hadn’t changed. The price changed, and the “medicine” followed it. Britain pays for illness and gets illness; America pays for treatment and gets treatment. Neither pays for health, so neither gets it.

SuperFreakonomics contains the miniature version of the fix. Cedars-Sinai hospital in Los Angeles could not get its doctors to wash their hands; compliance hovered around 65 per cent while the hospital begged, lectured and dangled Starbucks cards. What worked was a redesigned incentive. The administration cultured the palm prints of its own executives, photographed the horror-film colonies of bacteria that bloomed, and made the image the screensaver on every computer in the hospital. Compliance went to nearly 100 per cent. Nothing about the doctors changed. The feedback loop changed: the invisible cost of the behaviour was made visible, immediate and personal. Hold that shape in mind, an incentive redesign achieving what a decade of exhortation couldn’t. It is the entire argument of the next essay in one anecdote.

Pay people not to work, and tax the ones who do

Here is where the incentive rot stops being a fiscal problem and becomes an existential one. In February 2026 the Centre for Social Justice ran the arithmetic on the whole package: a jobless claimant combining Universal Credit’s health element, housing support and PIP can receive around £25,200 a year, more than the take-home pay of a worker on a £30,100 salary. On the CSJ’s numbers, 6.2 million British workers, roughly a quarter of all full-timers, take home less than they could receive for not working at all. Critics quibble with the aggregation, and fairly, because not every claimant gets the full stack. But the machine only needs to be roughly right to reshape behaviour. A society that pays its marginal worker more to exit the labour force than to stay in it has not made a moral error. It has posted a bounty on economic inactivity, and bounties get harvested.

For those who insist on working anyway, the system dangles a second set of disincentives at every rung of the ladder. Earn more and Universal Credit tapers away at 55p in the pound on top of tax and National Insurance, so millions of low earners keep less than 30p of each extra pound. Build a company and corporation tax takes 25 per cent of the profit, up from 19 in 2023, before dividend tax takes another slice of what’s left, while the 2024 Budget raised employer National Insurance and cut its threshold, a tax on each additional person you dare to hire. Climb high enough and you hit the £100,000 cliff, of which more in a moment, where the rational move is to ask your employer to pay you less. At every income level, from benefits claimant to consultant surgeon to company founder, the marginal unit of effort in Britain is taxed, tapered or clawed back. The country has assembled, with no one intending it, a complete end-to-end disincentive stack: a bounty for not working, a penalty for working more, and a penalty for employing anyone else.

Saw off the first rung of the ladder, and call it kindness

One more piece of the employment machine deserves holding up to the light, because it is the most counterintuitive and the most politically untouchable: the minimum wage. Nobody disputes the intention, which is decency itself. But look at the mechanism. A minimum wage is not a law that says employers must value workers more highly. No statute can do that. It is a law that says: if this person’s output is worth less than the floor, employing them is illegal. The worker whose value-add sits below the line doesn’t get a raise. They get nothing, because the only legal wage for them is zero.

Economists have argued for thirty years about how hard this bites, and at modest levels the honest answer is: less than the textbook says. But Britain is no longer running it at modest levels. The adult floor hits £12.71 in April 2026, among the highest ratios to median pay in the developed world. The 18-to-20 rate is being force-marched upward, 8.5 per cent in a single year, on an explicit path to abolishing the youth discount altogether. And stacked on top sits the employer National Insurance rise, which now bites from just £5,000 of salary, so the state collects its cut of a job almost from the first hour worked. Each measure is defensible alone. Together they are a steep and rising tax on the precise act of giving an inexperienced person their first chance.

The results are arriving on schedule. Hospitality, the classic first-job sector, shed 108,000 payrolled jobs in the year to July 2025, the worst of any UK sector, 84,000 of them after the NI rise was announced, then lost another 8,784 jobs in December 2025, in the middle of the festive season, when it should be hiring everything that moves. Youth unemployment reached 16.1 per cent in early 2026, the highest in over a decade. The teenager who would happily sweep floors, pull pints and learn to show up on time for £9 an hour is not permitted to make that trade at any wage an employer can now justify. So they make no trade at all, and drift toward the benefits system in the previous section, which, recall, pays better than the job they can’t get. The ladder’s first rung hasn’t been raised. It has been sawn off, with a plaque screwed on where it used to be, reading “dignity.”

Abolish the disincentive, and you’ve created an incentive

Crime is the same ledger with the sign flipped. A functioning justice system is a price list: steal this, and it will probably cost you that. Like any price, it only works if it’s actually charged. Stop charging it, and you haven’t merely gone soft on crime. You’ve changed the expected-value calculation of a life of theft from negative to positive, and posted yet another bounty.

Freakonomics readers already know the deep mechanism, because Levitt and Dubner made it famous with a study of Israeli daycare centres. A group of nurseries in Haifa, tired of parents arriving late for pickup, introduced a small fine, about three dollars. Late pickups didn’t fall. They roughly doubled, and stayed high even after the fine was scrapped. The fine had converted a moral obligation, don’t make the teacher stay late, into a commercial transaction with a posted price, and at three dollars, late pickup turned out to be a bargain. The economists Uri Gneezy and Aldo Rustichini titled their paper “A Fine Is a Price,” which is the whole lesson in five words: a penalty small enough becomes a tariff, and a tariff licenses the behaviour it taxes.

Britain ran the daycare experiment on crime itself. In 2014, Parliament classified shop theft under £200 as “low-value shoplifting,” a summary offence you could settle by post. The intent was court efficiency. The signal received on the street, and, by the government’s own later admission, inside police forces, was that theft under £200 was no longer real crime; ministers ended up describing the rule as an “effective immunity.” Retailers told their staff not to intervene, police attendance became a rumour, and the market responded exactly as it responds to every mispriced bounty in this essay. Police-recorded shoplifting climbed to around 530,000 offences a year by 2025, the highest since records began two decades ago, before easing a hair to 509,566 in the year to December 2025. The British Retail Consortium’s survey puts actual customer theft at over 20 million incidents a year, costing retailers £2.2 billion; set that against the recorded half-million and something like 96 per cent of shop theft never enters the statistics at all. It has simply been priced into the weekly shop of every honest customer, a private tax collected at the till.

The wider ledger says the same. Across all victim-based crime in England and Wales in 2024-25, just 6.3 per cent of recorded offences ended in a charge. For theft, 70.8 per cent of cases closed with no suspect identified at all. Walk out with someone’s phone, someone’s bike or an armful of someone’s stock, and the modal consequence is nothing, a probability any rational shoplifter prices in about as fast as a Hanoi rat farmer. Parliament finally repealed the £200 rule in April 2026, restoring a seven-year maximum for shop theft of any value, which is welcome, and also an admission: the state itself now accepts that it spent a decade running an incentive scheme for shoplifting. Britain had managed to build both halves simultaneously: negative marginal reward for effort, positive expected value for theft. No civilisation keeps that configuration for long.

The catalogue of madness

Once you have the lens, you can’t stop seeing it.

The UK tax system contains a band, between £100,000 and £125,140, where withdrawal of the personal allowance pushes the marginal rate to 62 per cent, and the loss of childcare support at £100,000 goes one better: the IFS calculated that a London parent with two nursery-age children earning £100k would need a pay rise to roughly £134,500 to be better off than at £99,000. A tax system with marginal rates above 100 per cent is the state paying its most productive people, at the peak of their careers, to work less. Surveys suggest four in five earners near the threshold oblige. Stamp duty taxes the act of moving house rather than owning one, so 40 per cent of English households under-occupy while young families can’t find homes, and the pensioners rattling around four-bedroom houses are behaving with perfect economic rationality. Until 2023 the pension lifetime allowance was literally paying senior NHS surgeons to retire early in the middle of a waiting-list crisis, which is why Jeremy Hunt abolished it, as an NHS retention measure.

America’s federal government guaranteed unlimited graduate student loans, so universities raised prices to absorb them; the New York Fed found sixty cents of every subsidised loan dollar passing straight into tuition, and student debt hit $1.66 trillion before Congress finally abolished Grad PLUS loans this month. Civil asset forfeiture lets police keep what they seize without a conviction, and they have seized at least $82 billion since 2000, with the median cash grab, $1,678, sitting conveniently below the cost of the lawyer you’d need to get it back. Federal flood insurance charges below-risk premiums and then pays to rebuild the same houses, including one Louisiana property worth $56,000 that has collected $183,000 across forty claims. The programme owes the Treasury $20.5 billion and counting.

Europe, never knowingly outdone, once paid farmers guaranteed prices for whatever they produced and duly warehoused a 1.2 million tonne butter mountain and a wine lake of a billion-plus surplus bottles a year, then pivoted to paying farmers per hectare, and at one point paid them to not farm 15 per cent of their land at all. Germany guaranteed solar producers above-market prices for twenty years regardless of whether anyone needed the power, and by 2025 German electricity prices went negative for 573 hours a year while consumers paid €435 million compensating operators for power that was never produced. And Italy, the reigning world champion, decided in 2020 to reimburse home renovations at 110 per cent of cost, in freely tradable tax credits. Sit with that design for a second: the state pays you more than the work costs, and you can sell the receipt. Every incentive to inflate the invoice, none to economise. Estimated cost €35 billion; actual cost around €220 billion, twelve per cent of GDP, with €15 billion of identified fraud and the finance minister describing his own country’s policy as “wicked.” Italy renovated its way into a fiscal crater at more than €1,000 per tonne of carbon abated, roughly twelve times the market price of simply buying the emissions reduction.

None of these are stories about evil people. Every single one is Hanoi: rational actors, faithfully harvesting the bounty a well-intentioned committee actually posted, rather than the outcome the committee imagined it was buying. The Soviets ran the pure laboratory version, where glass factories on tonnage quotas made glass too thick to see through and, switched to area quotas, made it too thin to survive delivery. Freakonomics assembled the definitive micro-catalogue. When Chicago attached high stakes to school test scores, Levitt’s forensic algorithm caught teachers outright changing pupils’ answers in around five per cent of classrooms; the incentive was test scores, so it bought test scores, by any means available. In sumo, where a wrestler’s eighth win in a fifteen-bout tournament is worth vastly more than his seventh loss, wrestlers sitting at 7-7 on the final day beat 8-6 opponents about eighty per cent of the time, against an expected fifty; the rank incentive was lopsided enough to visibly bend one of the most honour-bound sports on earth. And estate agents, who pocket only a sliver of each marginal pound of sale price, leave their own homes on the market about ten days longer and sell them for roughly three per cent more than their clients’; when the incentive to hold out for a better price is theirs rather than yours, mysteriously, they hold out. Goodhart’s law, as reformulated by Marilyn Strathern, covers every case in eleven words: when a measure becomes a target, it ceases to be a good measure.

Underneath all these stories sit the same three ingredients.

First, the people allocating the capital are not spending their own money, so the feedback loop from bad allocation to personal pain is severed. Milton Friedman’s fourth quadrant: other people’s money spent on other people, with predictable care taken over both amount and quality.

Second, value is defined upstream by a committee, once, in a document, instead of being discovered continuously by the people consuming it. The committee writes “dead rats” but pays on “tails,” and can’t update the contract until the next legislative session, by which time the rat farms have lobbyists.

Third, there is no mechanism for defunding failure. A subsidy that misfires doesn’t shrink; it acquires beneficiaries, and beneficiaries vote. The sickness benefit surge survived its own reform. The Superbonus ran for years after everyone knew. The Common Agricultural Policy still consumes a third of the EU budget forty years after the butter mountain.

Corporations are better than governments at this, but only by degree. A corporation is still a central planner internally: capital allocated by an annual budgeting process, political capture between divisions instead of between constituencies, pay tied to proxies (hours, headcount managed, revenue booked) that drift away from value creation exactly as Goodhart predicts. The difference between a government programme and a corporate division is that the corporation eventually runs out of other people’s money slightly faster.

So the trillion-dollar question: what would it look like to build an economic institution where the incentive cannot drift from the outcome? Where value is measured continuously rather than legislated once, where capital reallocates automatically toward whatever is working, and where nobody, no minister, no CFO, no root committee, holds the purse strings?

It isn’t hypothetical. We have had a working prototype for seventeen years. Bitcoin posts a bounty, currently around $10 billion a year, for exactly one commodity: valid proof-of-work on the honest chain. Not tails. Not proxies. The thing itself, verified by mathematics rather than by inspectors who can be fooled by a tailless rat. The result is the most secure computer network ever built, constructed by anonymous strangers with no CEO and no budget committee. Bitcoin took the exact human energy that farms rats, greed, and pointed it at a socially useful output by making the reward function ungameable.

And for the last few years, a network called Bittensor has been industrialising that trick: a factory for building Hanoi-proof bounties that can buy anything measurable, machine intelligence, forecasts, compute, and that reallocates its entire capital budget across a hundred-plus competing markets continuously, by price, with no allocation meeting, ever. What that means for companies, for investors, and eventually for the governments whose allocation record you have just read, is the subject of the next essay in this series.

The rat catchers were never the problem. They were magnificent. They always were. Change what you pay for, and you change the world.