Sixteen Experiments
There is a kind of accelerator that funds sixteen startups, taxes the revenue of the ones that ship, hands it to the ones that don’t, forbids any of them from competing on price, and writes the arrangement into a constitution nobody can amend.
That accelerator is Germany. The startups are the Länder. The constitution is the Basic Law.
The conventional story is that German federalism is one of the country’s quiet strengths. Sixteen states with real constitutional power. An eternity clause that cannot be amended away. A Bundesrat with teeth. By the textbook definition this is federalism. By the working definition, the one that explains why federalism is supposed to make a country stronger over time, it isn’t.
Federalism’s actual job is not to make a country more polite. It is to turn the country into a learning algorithm. Sixteen states with real power are sixteen experiments running in parallel. A state that sets smart tax rates, spends money well, and builds good infrastructure attracts people and companies. A state that doesn’t, bleeds them. The migration is not cruelty. It is information.
For the system to learn, two things have to be true. The experiments have to vary. The results have to be visible. Mute either signal and federalism stops working, no matter what the constitution says.
Germany has muted both.
The first severing is taxation. The federal government sets nearly every tax that matters. Corporate, income, VAT, capital gains. The states get a few token levers, and municipalities can tune the Gewerbesteuer multiplier, but none of it adds up to enough to change a business’s mind about where to land. A founder choosing between Munich and Stuttgart is not really choosing between tax regimes. They are choosing between the same tax regime in two cities.
The second severing is the Länderfinanzausgleich. Wealthier states transfer money to poorer ones to keep gleichwertige Lebensverhältnisse across the country. The scale is not symbolic. In 2023, Bavaria paid roughly €9.4 billion into the system. Berlin took in around €3.8 billion, Saxony another €3.2 billion. Bavaria has been a net payer for decades. Berlin has been a chronic net receiver. The numbers move every year but the direction does not. Performance at the state level is detached from reward. Underperformance is detached from cost.
Each severing by itself would weaken the loop. Together they short it out. A state can’t compete on tax because the levers are at the center. A state can’t win quietly by being well-run, because the surplus gets equalized away. The system that was supposed to learn from sixteen experiments runs sixteen experiments and then averages the results back into one.
I wrote in Hormesis that what doesn’t get loaded doesn’t get strong. Bone in zero gravity does not preserve itself. It weakens, because the body reads comfort as a signal that strength is no longer needed. The Länderfinanzausgleich is bone in zero gravity for entire states. Thirty-five years after reunification, the transfers are still flowing in roughly the same direction, and the gap they were meant to close has not closed. That is not a bridge. It is life support.
The fix doesn’t require anyone to be abandoned. Put a floor under the things that have to be irreversible: basic education, child welfare, a thin safety net. Above that floor, give states back the levers a business actually cares about. Tax rates first, but also regulation, administration, the speed of getting a permit. Time-limit the investment money that goes to states starting behind. Let the permanent income transfers expire. Solidarity that funds launchpads, not hammocks.
Tax competition is supposed to be a race to the bottom. States compete to be cheap, not good. The winner is whoever is most willing to gut their commons. It sounds like a law of nature when stated abstractly. It is an empirical claim, and the empirics disagree.
Switzerland’s twenty-six cantons set their own income, corporate, and wealth taxes. The variation is not cosmetic. Zug taxes corporate profit at around 11.8%; Geneva sits closer to 14%, having recently cut its rate to stay competitive. Top personal income tax in Zug runs about 22%; in Geneva it can exceed 40%. Decades into this regime, Switzerland is not a wasteland. It is among the richest countries on earth, with excellent public services, world-class education, a functioning healthcare system, and a quality of life wealthy Germans periodically vacation in to remember what one looks like. If tax competition produced the race to the bottom the theory predicts, we would not have to argue about Switzerland. We could point at the rubble. There is no rubble.
The reason the story doesn’t survive contact with Switzerland is that competition between states is multi-dimensional. Citizens don’t only care about taxes. They care about schools, transit, safety, the speed of the bureaucracy, the air. A state that cuts taxes by hollowing out the schools loses families. A state that cuts taxes by becoming better-run, faster, less wasteful, wins them. Apple does not win by being the cheapest phone. A well-run canton does not win by being the cheapest canton.
The American case is messier than Switzerland and more useful for that reason. Fifty states. Fifty different tax regimes, regulatory regimes, school systems. The signal is loud, sometimes ugly, and unmistakable.
For most of the last decade, the largest net flow of domestic migration in the United States has been out of California and into Texas and Florida. Between 2020 and 2023, California lost roughly 1.2 million residents net. The reasons are not mysterious. California taxes more, regulates more, builds less housing. Texas taxes less, regulates less, builds aggressively. People with the choice are voting on the difference, in numbers visible from orbit. And not the people the critics assume. The Black population of San Francisco has fallen from 13.4% in 1970 to roughly 5% today. The families being priced out are disproportionately the working-class and minority families the regulation was supposed to protect. Companies are following the people. Capital is following the companies. California is not collapsing. But it is being read, in real time, and the read is being acted on.
Pick any other dimension and the same thing is happening. Massachusetts school reforms get studied and copied. Some of the copies stick, some don’t, and the country gets quietly smarter. None of this requires a central planner to identify best practices. The states identify them by being out-competed on them. The system learns because it has the property federalism is supposed to give it, and which Germany has spent the postwar period systematically removing: a signal you can read, and a consequence you can feel.
The strongest objection is mobility. Capital moves frictionlessly. Corporations relocate in a quarter. But poorer people, the elderly, families anchored in their communities, often cannot. So the signal gets dominated by the most mobile actors. Policy bends to please the rich. The market for governments turns out to be a market that serves capital and abandons the rooted.
The signal does not require mobility to function. A citizen who cannot move can still see that the next state has better schools and lower taxes and a more responsive administration. They vote in their own state accordingly. Feet are one channel. Ballots are another, and ballots do not require a moving truck. The fully immobile voter in a federation where the signal works is a voter who can point at a working example next door and demand it.
The floor exists precisely for this. A national minimum on the irreversible things is what protects the immobile from being abandoned. You do not have to choose between total equalization and total abandonment. The actual choice is between floor-plus-autonomy and ceiling-plus-rescue, and Germany has been doing the second for so long that the first has become unimaginable.
And the current system does not deliver on its own promise. The transfers have not closed the gap they were built to close. The immobile in Mecklenburg-Vorpommern have been told for thirty-five years that the transfers are why their schools and infrastructure and prospects will eventually look like Bavaria’s. They do not look like Bavaria’s. The transfers have funded the gap, not closed it. The people the system claims to serve have started to notice, and the election maps have started to show it.
The eternity clause was supposed to lock federalism into the country forever. It locked the form in. Then the country spent the postwar period quietly amending what it was supposed to be guarding. The states have their constitutional power. They no longer have the levers that power was for.
Federalism, properly built, is not a polite arrangement among equal partners. It is a competitive machine that gets smarter by running. Sixteen experiments. Open results. Real consequences. A country that improves not because anyone at the center became wise, but because the system was finally allowed to learn.
Sixteen startups are still running. The accelerator has just stopped keeping score.