The Chess Pieces Moved

In 1973, Ingvar Kamprad left Sweden. He had founded IKEA. Sweden had a wealth tax. He moved to Denmark, then to Switzerland three years later. Hans Rausing, whose family built the Tetra Pak empire, left for England in the early 1980s. By the time Sweden abolished the tax in 2007, the country's own tax authority estimated that 500 billion kronor in Swedish capital sat offshore for tax reasons. The wealth tax had never exceeded 0.4 percent of GDP.¹

France created its wealth tax in 1982. Over the following thirty-five years, sixty thousand millionaires left the country. The economist Eric Pichet calculated that the tax cost France roughly twice what it raised in revenue. In 2018, Emmanuel Macron replaced it with a tax on real estate only. The logic was simple. You cannot move your building to Switzerland.²

In September 2022, Kjell Inge Rokke, Norway's third-richest man, published an open letter in the country's largest financial newspaper. Norway had just raised its wealth tax rate from 0.85 to 1.1 percent. "A difficult choice has been made," he wrote. "I have moved from Asker, Norway, to Lugano in Switzerland." More than thirty wealthy Norwegians left that year. More than the previous thirteen years combined.³

The belief, in each case, was that the tax would raise revenue because the people being taxed would stay where they were. They did not stay. They responded to the changed incentive exactly as an economist would predict and exactly as the policymakers did not.

Nobody abolished theirs because they were ideologically opposed to taxing wealth. Sweden, Denmark, and France are among the most redistributive governments on earth. They abolished the tax because the people they were taxing left.

The three that survived share a structural feature the nine that abolished did not. Switzerland's wealth tax is cantonal, not national. Different cantons charge different rates. The wealthy move between cantons, not out of the country. And Switzerland is where everyone else's millionaires flee to. There is nowhere better to go. Norway's threshold is roughly $264,000. Seven hundred and twenty thousand people pay it out of 5.5 million inhabitants. You are not taxing a small group of people who can afford to hire a lawyer and leave. You are taxing a significant share of the population. Spain introduced its national version in 2022 and made it permanent. The jury is still out.⁴

The nine that abolished had taxed a narrow group at high rates. A small number of very wealthy people with maximum incentive and maximum ability to leave. The tax was designed as if those people had no alternatives. They had the most alternatives of anyone in the country.

In 2008, Maryland ran the same experiment at the state level. A surcharge on residents earning more than one million dollars. The top state income tax rate rose from 4.75 to 6.25 percent.⁵

The financial crisis contributed. Capital gains collapsed everywhere that year. But the filers did not simply earn less. Hundreds of them stopped filing in Maryland altogether. The state had modeled them as a line on a spreadsheet. They turned out to be people with accountants, second homes, and options.

Maryland's surcharge was temporary. It expired in 2010. What followed was not.

Washington State ran a version of the experiment in 2021. The state had never had an income tax. A 7 percent tax on capital gains exceeding $250,000 was the closest thing anyone had managed to pass. In its first year, it collected $840 million. The top ten payments accounted for $394 million of the total. Half the revenue from ten people.⁶

In its second year, collections fell to $419 million. The top ten payments fell to $142 million. The tax had not changed. The composition of who was paying it had.

In November 2023, Jeff Bezos announced he was moving from Seattle to Miami. He cited his parents. He sold $13.6 billion in Amazon stock the following year. Fortune estimated his savings from Florida residency at roughly one billion dollars for 2024 alone.⁷

In March 2026, Washington's legislature passed a 9.9 percent tax on income exceeding one million dollars. The state's first income tax in ninety-three years. The House debate ran twenty-five hours. The final vote was 51 to 46. During the twenty-third hour, Howard Schultz posted on LinkedIn that he was leaving Seattle. He had built Starbucks there over four decades. He had purchased a penthouse in Surfside, Florida for $44 million. "It is our hope," he wrote, "that Washington will remain a place for business and entrepreneurship to thrive."⁸

The tax is projected to raise $3.5 billion annually from roughly thirty thousand taxpayers. It takes effect in January 2028. The departures began before the governor signed it.

In California, the departures began before the proposal qualified for a ballot. In October 2025, a union-backed coalition filed an initiative for a one-time 5 percent tax on residents with a net worth exceeding one billion dollars. The initiative needs 875,000 signatures. By March 2026, it had collected a quarter of them. The tax obligation date was set for January 1, 2026.⁹

On December 31, 2025, Peter Thiel announced a new office in Miami. David Sacks announced one in Austin. Both billionaires. Both on the last day of the year. A financial adviser told Bloomberg he had personally helped four billionaires relocate before year-end. Sergey Brin committed $20 million to defeat the measure.¹⁰

The initiative is not law. It is a proposal, backed by a fraction of the signatures required to reach the ballot. The billionaires did not wait for the vote. The chess pieces moved before the hand reached the board.

The pattern

In 2016, Wells Fargo's employees had been opening fraudulent accounts for fourteen years. The bank's cross-selling program, internally called "Going for Gr-Eight," set a target of eight products per customer. It rhymed with "great." The number was not a suggestion. Employees who missed it faced consequences. So they opened accounts that customers had not requested. Roughly 3.5 million of them.¹¹

The original fine was $185 million. The total penalties, over the next decade, exceeded $5 billion. The Federal Reserve imposed an asset cap in 2018 that was not lifted until 2025. The measure had become the target. The target had consumed the institution.

Apple charged app developers a 30 percent commission for in-app purchases. In 2020, Epic Games deliberately bypassed the payment system. Apple removed Fortnite. Epic filed suit that afternoon. In 2021, a federal judge ruled that Apple's anti-steering provisions violated California law and ordered Apple to let developers link to external payment options. Apple complied by charging a 27 percent commission on those external purchases. In April 2025, the same judge found Apple in willful contempt and referred the company to prosecutors. Days earlier, the European Commission had fined Apple 500 million euros for violating its own anti-steering rules. In December, the Ninth Circuit affirmed the contempt finding and sent the case back to determine what commission Apple could charge.¹²

Each attempt to hold the 30 percent produced a more aggressive counter-response. Developers routed around. Regulators intervened. Courts escalated. The commission was designed as a wall. It became a provocation.

In September 2024, Amazon ordered every employee back to the office five days a week. JPMorgan Chase followed in January 2025. Dell began tracking badge swipes and VPN usage, assigning employees color-coded flags in its HR system. Blue for consistent. Red for limited. Remote workers were told they would no longer be eligible for promotions.¹³

A University of Pittsburgh study tracked more than three million tech and finance workers on LinkedIn across fifty-four S&P 500 firms. After return-to-office mandates, the departure rate rose 14 percent. The increase was not uniform.¹⁴

Gartner surveyed 2,080 knowledge workers and found that high performers' intent to stay dropped 16 percent under strict mandates. Double the rate of average employees.¹⁵

The best employees had options. They used them. The mandate was designed to increase presence. It changed the composition of who was present.

The four names

In April 1902, the French colonial administration in Hanoi had a rat problem. The city's new sewer system, a monument to modern sanitation, had become a habitat. The Third Plague Pandemic was moving through Asian port cities. The authorities needed fewer rats. They offered a bounty: one cent per rat tail. Whole carcasses were deemed unseemly for government buildings, so only the tails were required.¹⁶

Tails poured in by the thousands. Then officials noticed something. Rats were running through the streets of Hanoi without tails. People had been catching the rats, cutting the tails, and releasing the animals to breed. Some imported tails from the countryside. Some set up farms.

The bounty was designed to reduce the rat population. It created a rat-breeding industry. The program was cancelled. The breeders released their stock. The population rose above where it had been before the bounty existed.

Everyone calls this the cobra effect, after a story about a similar bounty on cobras in British India. That story is almost certainly apocryphal. No historian has found a date, a name, or an archival record. The Hanoi rat hunt is documented. Michael Vann and Liz Clarke traced it through French colonial archives and published it through Oxford University Press in 2018. The story that became the metaphor was never verified. The one that was verified is worse.¹⁷

The rat bounty does not need a theorist. The story is the law. You change the incentive. The targets adapt. The adaptation produces the opposite of your intent.

Wells Fargo set a target of eight products per customer and got 3.5 million accounts that nobody requested. Charles Goodhart, a Bank of England economist, described this in 1975 at a central banking conference in Sydney. Central banks had adopted monetary targets. Each chose the aggregate with the most stable historical relationship to inflation. By the time the targets were set, the relationships had broken down. "Any observed statistical regularity," Goodhart wrote, "will tend to collapse once pressure is placed upon it for control purposes."¹⁸

The social anthropologist Marilyn Strathern rephrased it in 1997: "When a measure becomes a target, it ceases to be a good measure."¹⁹

Norway, Sweden, and France set wealth as the target. Twelve countries ran the experiment. Nine reached the same conclusion. Charlie Munger, Berkshire Hathaway's late vice chairman, put it more directly: "Show me the incentive and I will show you the outcome."²⁰

Thomas Sowell arrived at the same observation through different evidence. In the early 2010s, a wave of American cities and states adopted "ban the box" legislation. The laws removed criminal history questions from job applications. The intent was to help people with records get past the initial screen. The assumption was that once employers met the applicant, the record would matter less.

In 2018, Amanda Agan and Sonja Starr published the results of a field experiment in the Quarterly Journal of Economics. They had sent roughly 15,000 fictitious applications across New Jersey and New York City, before and after ban-the-box policies took effect.²¹

Employers who could no longer see criminal records used race as a proxy for criminality. A separate study found that ban-the-box policies decreased the probability of employment by 3.4 percentage points for young, low-skilled Black men. The intervention designed to reduce discrimination in hiring had increased it.²²

Sowell had a term for this kind of error. He called it the chess pieces fallacy. The reference is to Adam Smith's observation that the "man of system" arranges the members of society "with as much ease as the hand arranges the different pieces upon a chess-board." The planner forgets that the chess pieces have "a principle of motion of their own."²³

Four descriptions of the same law. Goodhart observed it in monetary policy. Strathern generalized it to measurement. Munger compressed it to incentives. Sowell traced it through policy interventions across centuries. The Hanoi rats demonstrated it before anyone named it. Change the rules, and the people inside the system will change their behavior. The behavior change will not be the one you designed for.

The exception

Not every intervention produces its opposite.

In May 2023, Netflix enforced its password-sharing crackdown across the United States. The company had 232 million subscribers. Analysts expected 1.77 million additions that quarter. Netflix added 5.89 million. By the end of 2024, it had 301 million paid memberships. By the end of 2025, 325 million.²⁴

The chess pieces moved exactly as Netflix predicted. Password sharers were not casual users who would walk away. They were engaged viewers who watched regularly and had no comparable alternative. Netflix modeled them as agents with specific motivations, not as inert objects on a board. The prediction was correct because the assumption was correct.

An NBER working paper by Katrine Jakobsen, Henrik Kleven, and co-authors used decades of Swedish and Danish administrative data to measure the aggregate economic impact of wealth tax migration. They found what the individual stories obscure. Yes, a 1 percentage point increase in the wealth tax rate decreases the stock of wealthy taxpayers by about 2 percent. But the aggregate effects on the broader economy are modest. Employment declines 0.02 percent. Investment, 0.07 percent. "For each additional dollar of revenue raised by the wealth tax," they wrote, "only 0.22 dollars are lost through migration responses."²⁵

Norway's total wealth tax revenue rose from 27 billion kroner in 2022 to 34 billion in 2025, despite the departures. The tax applies to roughly 720,000 taxpayers, not just the ones who made the newspaper. Rokke left. The tax base held.²⁶

The law operates on a condition. The targets must have options. Millionaires have options. IKEA founders have options. High-performing engineers at Amazon have options. The people who can move, move. The people who cannot, stay.

This is what separates the chess pieces fallacy from a general theory of policy failure. Not every intervention backfires. The ones that backfire share a specific error. They model the targets as fixed. They assume the millionaire will stay in Maryland, the developer will keep paying 30 percent, the high performer will return to the office, the employer will evaluate the application the same way with less information. In every case, the target had an alternative the planner did not account for. In every case, the target found it.

Twelve countries designed a tax that assumed wealth would sit still. The wealth moved. Each country, in its own time, arrived at the same conclusion. Not because the tax was unjust. Because the assumption underneath it was wrong.

Goodhart saw it in central banks. Munger saw it in boardrooms. Sowell saw it in labor markets. A French colonial administrator in Hanoi saw it in the rats running through his city without tails. Four traditions, a century apart, describing the same structural error. The targets are not objects. They respond. And the response, more often than the planner expects, produces the precise outcome the intervention was built to prevent.

The chess pieces moved. They always do. The question was never whether they would move. It was whether anyone designing the board had considered that they could.