Haute Lumière

Commerce · VI.08 · MMXXVI · daylight

La Bourse  /  Volume VI  /  Nº VI.08

Federation and Subsidiarity

Volume VI — Governance and the Commons

---

THE PLATE

---

THE LETTER

You have been handed a question that sounds political and is not. Somebody wants a function moved — purchasing pulled up to the group, hiring pushed down to the site, safety standards written once for everybody, a brand held centrally and a menu held locally. The arguments on both sides are sincere and neither side has a number.

That is the whole difficulty, and it is fixable in an afternoon.

There is a real theory of assignment. It has been worked since 1972, it is taught in every public finance department, and almost nobody outside that department has met it — which is remarkable, because the question it answers is the single most common structural question in organisational life. Wallace Oates wrote down the condition under which a function belongs low. Charles Tiebout wrote down the mechanism that would make low assignment self-correcting, and listed the conditions it needs. Mancur Olson wrote down, in six pages, the rule that decides the rest: the boundary of the jurisdiction should match the boundary of the benefit. Put those three together and the argument in your meeting becomes an inequality with two terms in it, both of which you can estimate before lunch.

This chapter gives you that inequality, computes it, and then tests it against four federations that have been running it with real money for decades: the European Union's subsidiarity procedure, Switzerland's cantons, Germany's equalisation between the Länder, and a federation of cooperative banks with no state in it anywhere.

Chapter I.10 did the other half of this — where a mesh of peers stops paying for itself and what changing shape does to that ceiling. It answered how many. This one answers which function, at what level. Read them as one instrument if you can; they were cut to fit.

One promise about the honest part. Subsidiarity says a function goes to the lowest level capable of performing it, and capability is not evenly distributed. The arithmetic of that is uncomfortable and it is in here, measured against a real federation, with the price of the correction attached.

— The Editors

---

DISCOVERY

What is already working

Start where the assignment problem has been solved well, because it has been, repeatedly, and mostly by people who never read the theory.

Switzerland has been running the experiment for longer than anyone. Twenty- six cantons, and beneath them communes — 2,899 of them on the first of January 2000, and 2,131 on the first of January 2024. That is 768 communes merged away in twenty-four years, a fall of 26.5 percent, at a mean of 32.0 mergers a year, and not one of them was ordered from Bern. The communes merged when the functions they held outgrew them, and they did it by vote. A federation in which the units can reshape themselves is doing something no organisation chart can do: it is letting the boundary follow the function.

What Switzerland shows most clearly is that assignment has a visible price and the country pays it willingly. The effective combined corporate income tax rate runs at about 11.9 percent in the lowest canton and about 21.0 percent in the highest — a factor of 1.76 between two places a train ride apart. In most countries that spread would be an emergency. In Switzerland it is the constitution working: Article 3 reserves to the cantons everything not assigned to the Confederation, Articles 5a and 43a write subsidiarity in by name, and the tax spread is what a real reservation of powers looks like from the outside.

The European Union wrote subsidiarity into treaty law and gave it a procedure. Article 5(3) of the Treaty on European Union says the Union acts only where the objectives cannot be sufficiently achieved by the member states. Protocol No. 2 turns that into machinery: every legislative proposal goes to every national parliament, each parliament holds two votes, and they have eight weeks to object. It is the only place on earth where a philosophical principle about the right level of government has been given a vote count, a clock, and a published record. Whatever the record shows — and we will read it honestly in a moment — the fact that there is a record at all is the achievement.

Germany did the thing federations usually refuse to do: it made the transfer explicit. The Länder equalisation moved roughly 18.5 billion euros in 2023, Bavaria paid about 9.0 billion of it — 48.6 percent of the whole — and Berlin received about 3.6 billion. Per head that is about 672 euros paid by every Bavarian and about 947 euros received by every Berliner, a ratio of 1.41. Those numbers are published, arguable, and litigated; Bavaria has taken the system to the constitutional court more than once. A transfer that is argued about in public every year is a transfer that survives, which is more than can be said for the ones that are hidden in an allocation formula.

And a federation with no state in it at all. The German cooperative banks — the Volksbanken and Raiffeisenbanken — run a joint protection scheme that has been in continuous operation since 1934. In the ninety years to 2024 the number of member insolvencies that cost a depositor money is zero. That is not a claim about German banking; it is a claim about assignment. The members kept everything local that could be local — the lending decision, the customer, the branch, the board — and assigned upward exactly one thing: the risk pool. One function, correctly placed, held for ninety years.

Four federations, four continents of practice, one pattern underneath them: each of them got good by naming functions one at a time and placing each one separately. None of them decided, once, how centralised to be. That decision does not exist. There is only a list of functions, and each function has an answer.

---

THE ARITHMETIC

What works, what does not, and where the line sits

First, the theorem, stated with its assumptions, because the assumptions are where the whole argument lives.

Oates's decentralisation theorem (1972) says: in the absence of cost savings from central provision and in the absence of inter-jurisdictional spillovers, welfare is always at least as high — and generally higher — if each jurisdiction provides the Pareto-efficient level for its own residents than if a single uniform level is imposed everywhere.

Read the three conditions it needs, in order, because each one is a real constraint on the theorem's reach:

1. No spillovers. Benefits stop at the border.
2. No economies of scale in central provision.
3. The centre cannot differentiate — it must set one level for everybody.

The third is the one that has moved since 1972. Besley and Coate (2003) showed that if the centre can vary provision across regions, the uniformity cost collapses and the case for devolution has to be rebuilt on political-economy grounds rather than on the theorem. Most modern central governments can differentiate. Most group functions in a firm can too. So the theorem is not a licence to devolve. It is a statement about what devolution buys, priced in preference heterogeneity, and it holds only when the centre is genuinely constrained to a single answer.

Second, the mechanism the theorem leans on, and the conditions it needs.

Tiebout (1956) supplies the correction mechanism: if people can move, they sort into the jurisdiction whose bundle of services and taxes suits them, and local provision becomes efficient without anybody having to reveal a preference. It is a beautiful argument. It needs seven conditions, and the ones that fail are measurable.

Take mobility first. In the United States Census mobility series, 7.8 percent of the population moved in 2023–24. Of movers, about 64 percent stay inside the same county — so the share who actually cross a jurisdiction in a year is:

  movers                                  7.8%
  x share crossing a county line         36.0%
  ----------------------------------------------
  jurisdiction-crossers per year        2.808%

At that rate, the time for half a mis-sorted population to have re-sorted is ln(0.5) / ln(1 − 0.02808) = 24.3 years. In 1948, when the same series recorded 20.2 percent movers, the half-life was 9.2 years. Tiebout's mechanism has become 2.65 times slower within living memory, and it now clears at about the length of a human generation. A budget cycle is four years. A chief executive's tenure is shorter. The mechanism is real and it is not available to anybody making a decision this year.

Two more conditions, measured the same way. Tiebout's fourth assumption is that income does not depend on where you live; the mean one-way commute in the United States is 26.8 minutes, or 223.3 hours a year, which is the price of pretending otherwise. His third assumption is a large number of communities to choose from; there are 90,837 local governments in the United States, of which 19,479 are general-purpose municipalities — across 3,031 counties, that is 6.4 municipal choices per county. The large number is a national fact. The menu in front of any actual household is six.

And then the finding that should be better known than it is. Rhode and Strumpf (2003) tracked heterogeneity across American local jurisdictions from 1850 to 1990, through the largest fall in mobility cost in human history. If Tiebout sorting dominated, jurisdictions should have become more unlike each other. They became less unlike. Sorting is present; it is not the main force.

Third — the rule. This is the instrument, and it is the reason for the chapter.

Two losses, both measured as a fraction of the same thing: W*, the annual value at stake in getting this one function right.

The loss of assigning up is Oates's uniformity cost. Let h be the coefficient of variation of the ideal provision levels across the units — the standard deviation of what each unit would choose, divided by the mean. With quadratic welfare loss, the average loss from one uniform level is h².

The loss of assigning down is the spillover cost. Let s be the fraction of the benefit that lands outside the deciding unit. A local government internalises only (1 − s) of the benefit, so it stops short of the efficient level. With linear marginal benefit and constant marginal cost, the shortfall as a fraction of the efficient level is ρ·s/(1 − s), where ρ is a curvature term the reader sets — ρ = 1 puts the efficient level at the midpoint of the demand curve and is the neutral default. The loss is that shortfall squared.

  assign UP   costs   h²
  assign DOWN costs   ( ρ·s/(1 − s) )²

  they are equal when      s* = h / (ρ + h)
  and with ρ = 1           s* = h / (1 + h)

W* cancels. The rule needs no money in it to be applied, which is why it can be used on a function nobody has costed.

Now look at what it says, because this is the cut:

  h      s*      read as
  0.10   9.1%    units differ by a tenth  -> up past 9.1% leakage
  0.20   16.7%   units differ by a fifth  -> up past 16.7% leakage
  0.30   23.1%
  0.40   28.6%
  0.50   33.3%   units differ by HALF     -> up past 33.3% leakage

Intuition says: keep it local unless it obviously spills. The arithmetic says the crossover sits far lower than that. For units whose preferences differ by a quarter — already a heterogeneous federation — the threshold is 20.0 percent leakage. And for units that differ by half, which is about as unlike as members of one federation ever are, the threshold is still only 33.3 percent.

Subsidiarity's own arithmetic is more centralising than subsidiarity's rhetoric. The reason is structural and worth holding: the spillover term is convex in s — leakage hurts at an accelerating rate — while the uniformity term is merely quadratic in h, which is a bounded quantity, because preferences cannot differ by more than everything. A principle argued in words over-devolves, every time, and the over-devolution is invisible because the cost shows up as under-provision rather than as a line item.

Worked once, in full:

  spillover fraction s                     0.35
  heterogeneity h                          0.28
  shortfall  1 x 0.35 / 0.65             0.5385
  loss of assigning DOWN  = 0.5385²      0.2899
  loss of assigning UP    = 0.28²        0.0784
  threshold s* = 0.28 / 1.28             0.2188
  ratio of the two losses                  3.70

The function goes up, and it goes up by a factor of 3.70, which is a sentence that can go in a board paper.

Six real functions, assigned:

Schooling's spillover is not a guess: about 57 percent of Americans live in the state they were born in, so roughly 0.43 of the benefit of educating a child lands somewhere else. Every federation on earth has already moved education upward, and the rule says why.

Look at land use, because it is the row that fights. Real leakage — a fifth of the consequence of a planning decision lands on somebody else — and yet it stays local, because preferences about the shape of a street are so unlike that uniformity costs more than the leakage does. That is not a failure of the rule. That is the rule telling you where federations will always argue, and it is where they do.

Fourth: the record of subsidiarity as actually applied.

The European Union's early warning system has run since December 2009. To December 2025 that is sixteen years. In those sixteen years it has produced three yellow cards.

The mean margin is 2.67 votes. A unicameral parliament casts two. Every yellow card in the history of the procedure passed by less than two national parliaments' worth of votes. The rate is one card every 5.33 years. The orange card — a simple majority, 29 votes in an EU of twenty-eight — has never been triggered; the highest vote ever cast — twenty-two — is 75.9 percent of that threshold. One of the three proposals was withdrawn, which is 33.3 percent. And the number of Union acts the Court of Justice has annulled for breach of subsidiarity is zero.

None of that means the procedure is idle. Hundreds of reasoned opinions have been written, and the discipline of having to justify the level in every proposal is real. But read as an instrument, the conclusion is exact: a principle with a procedure and no arithmetic behind it fires once every five years and changes one proposal in three. The rule above is what the procedure is missing, and it would fit on the first page of a proposal.

Fifth, and this is the honest negative.

Subsidiarity assigns a function to the lowest capable level. Capability is income. So strict subsidiarity, applied without correction, systematically advantages the localities that were already rich — and the size of that advantage is measurable.

Switzerland publishes a resource index, rebased to a national mean of 100. The strongest canton runs above 250. The weakest sits near 67. That is a raw spread of 3.73 to one in the capacity to fund a function from own resources. A unit that can fund 67 percent of the efficient level loses (1 − 0.67)² of the function's value — 10.89 percent, gone, not to waste but to poverty, in the canton that could least afford it.

Switzerland corrects this with a legal floor. Since 2020 the equalisation system guarantees that the weakest canton reaches 86.5 on the index. At the floor, the capability loss is (1 − 0.865)² = 1.82 percent. The floor removes 9.07 percentage points of the function's value, and the whole equalisation apparatus costs about 5.5 billion Swiss francs a year against a GDP of about 801 billion — 0.687 percent of GDP.

Germany corrects the same thing differently and almost completely. Its pre-equalisation capacity index runs from about 120 down to about 78 — a spread of 1.54, far narrower than Switzerland's — and after equalisation the weakest Land sits at about 99.0. Capability loss falls from 4.84 percent to 0.01 percent, a removal of 4.83 percentage points, for 18.5 billion euros against a GDP of 4,185 billion: 0.442 percent of GDP.

Now put the two designs side by side, per unit of GDP spent:

  Switzerland   removes 9.07 pp   for 0.687% of GDP   =  13.21 pp per 1% of GDP
  Germany       removes 4.83 pp   for 0.442% of GDP   =  10.93 pp per 1% of GDP
  Switzerland's design returns 1.21x

And the reason, which is the usable part: the loss is quadratic, so it is concentrated at the bottom. Lifting the weakest Swiss canton from 67 to the floor of 86.5 is 19.5 index points and removes 9.07 percentage points — 0.4650 points of loss per index point. Lifting it further, from 86.5 to 99.0, is 12.5 index points and removes only 1.813 — 0.1450 per index point. The first stretch of the journey is worth 3.21 times the second.

So the honest sentence has both halves in it. Strict subsidiarity without equalisation hands the strong unit a structural advantage worth about eleven percent of every devolved function's value — and full equalisation is not the answer, because the last third of the levelling costs the most and buys the least. A floor is the answer, and the floor has a computable height. That computation is in Operationalize This, where it belongs.

---

DREAM

What becomes ordinary

In the federation that has done this, nobody argues about centralisation.

They argue about functions, one at a time, and the argument has a shape. Someone proposes that a function move. Two numbers arrive with the proposal: an estimate of how much of its benefit lands outside the deciding unit, and an estimate of how unlike the units' own answers would be. The estimates are rough and everyone says so. They are still enough, because the threshold h/(1 + h) is not a close call for most functions — the six-function table above has one row where the answer is genuinely near the line and five where it is not.

The assignment register is a document. It lists every function the federation performs, the level it sits at, the two parameters, the date they were last estimated, and the name of the person who estimated them. It is reviewed once a year, and the review is short because most rows do not move. When a row does move, it moves because a parameter changed — a new pollutant crossed a watershed, a new technology let the centre differentiate where it could not before, a member grew capable of something it could not do last year — and the minute records which parameter and by how much.

The equalisation is a published floor, not a negotiated grant. Every member knows the index, knows the floor, and knows the formula that sets the floor's height. Nobody lobbies for it, because there is nothing to lobby: the floor is where the last pound of top-up returns exactly one pound of avoided loss, and that is arithmetic. The strong members pay it without resentment, partly because it is small — in the worked federation later in this chapter, the strongest member pays 1.60 percent of its own stake in the function — and mostly because the alternative is a federation whose weakest quarter cannot perform the thing they all depend on.

Exit is cheap and nobody uses it. The functions held centrally are held because the arithmetic put them there and everyone has seen the arithmetic, so leaving would mean paying the spillover cost yourself. Membership is not loyalty. It is a calculation each member has personally checked.

And the units do not all look the same, which is the visible sign that it is working. Different opening hours, different buildings, different tone, different prices where prices are local. A federation whose members are indistinguishable has assigned too much upward and is paying for it in a currency its accounts do not have a column for.

---

DESIGN

The structure that gets there

First, make the list. Not an org chart — a list of functions. Twenty to sixty lines for most organisations, one line per thing the federation actually does. Purchasing. Brand. Pricing. Hiring. Safety standards. Compliance. Product specification. Customer data. Treasury. Training. The list itself does most of the work, because the argument about centralisation is almost always an argument in which the two sides are holding different functions.

Second, estimate the two parameters, badly, on purpose. For each function, s and h, each to one decimal place, each by a named person in ten minutes. The estimates are bad and they do not need to be good; the threshold is a factor-of-two decision and the estimates are within a factor of two. Three routes to s that work:

• The leakage question. If this unit does this well, what fraction of the benefit is enjoyed by somebody who did not pay for it? Count neighbouring units, future members, and the federation's shared name.
• The migration route. For anything invested in people, s is roughly the fraction who leave. Payroll records answer this in an hour.
• The complaint route. Who complains when this is done badly, and are they inside or outside the unit? The share of complaints from outside is a serviceable s.

And two routes to h: ask five units what level they would choose and take the coefficient of variation directly; or, where the function is already local, take the actual variation in current provision, which is h revealed rather than stated.

Third, apply the rule and record the near-misses separately. Compute s for each row. Where s exceeds s by less than a third, mark the row contested and leave it where it is. A federation that moves its near-misses spends its whole political capital on the rows where the answer barely matters. Move the clear ones; leave the close ones alone; revisit in a year.

Fourth, separate assignment from execution. The rule says at what level a function should be decided. It says nothing about where it should be done. A standard set centrally can be executed locally; a decision held locally can be executed by a shared service. Olson's fiscal equivalence is about the boundary of the benefit, not the location of the staff, and conflating the two is how a correct assignment turns into an unpopular head office.

Fifth, build the floor before you need it. Measure each member's capability on one index, publish it, and set the floor by the formula in the next movement. Do this while the spread is small, because a floor introduced when a member is already failing reads as a rescue and is resented; the same floor introduced early reads as insurance and is bought.

Sixth, give the members a way to reshape. Switzerland's communes merged themselves — 768 of them in twenty-four years — because the law let them and nobody forced them. Frey and Eichenberger's proposal of functional, overlapping, competing jurisdictions is the same instinct made general: let a boundary form around a function rather than forcing every function through one boundary. Two members who want to share a back office should not need permission. A federation that can only change shape from the centre will keep exactly the shape it had when it was smallest.

---

DESTINY

How it holds when nobody is pushing

Three things keep an assignment honest, and each of them is cheap.

The register is reviewed on a date, not on a trigger. Anything reviewed by exception is reviewed when someone is angry, and an angry review moves functions for reasons the rule cannot see. A short annual review of a mostly-unchanging list costs an afternoon and defuses most of the year's arguments before they start.

The parameters carry a name and a date. An estimate with nobody's name on it cannot be challenged, so it hardens into a fact. An estimate with a name on it gets corrected by the person who knows better, which is the only mechanism that keeps the register true.

The floor is a formula, not a budget line. A budget line is renegotiated annually by whoever is strongest that year. A formula with published inputs recomputes itself, and the negotiation is about the inputs, which is a much better argument to be having.

Now the failure modes, named plainly, because every one of them is on the record.

Capability changes and the assignment does not follow. This is the big one, and it runs in the direction nobody expects. Rabobank's local member banks numbered 174 in 2010 and 106 in 2015 — a fall of 39.1 percent — and in 2016 the remaining 106 merged into a single legal entity. The driver was not ambition at the centre. It was that supervision, capital rules and compliance had raised the cost of being capable, and the lowest capable level had moved up underneath the federation while the federation's structure stayed where it was. The German cooperative banks show the same force over a longer window: about 7,100 of them around 1970, 672 at the end of 2024 — a fall of 90.5 percent over 54 years, a compound rate of 4.27 percent a year, every year, for more than half a century. Subsidiarity is not a settlement. It is a moving answer to a moving question, and the thing that moves fastest is what the lowest level is capable of.

Note the other half of that record, because it is the hopeful half and it is equally true: through a 90.5 percent consolidation of the members, the functions assigned to the federation did not move. The risk pool stayed at the federation level, where it had been since 1934, and in the ninety years to 2024 the number of member failures that cost a depositor money is zero. The members changed shape completely and the assignment held, because the assignment had been made about the function rather than about the members.

The centre learns to differentiate and nobody notices. Oates's theorem needs a centre that must give one answer. The moment the centre can vary its answer by region — and software has made that nearly free — the uniformity cost falls and several rows on the register should move up. Almost no federation re-reads its register after acquiring that capability.

The floor becomes a grant. A floor tied to a published index is arithmetic. A floor that gets topped up "just this year" for a member in trouble is a grant, and grants are permanent. Keep the two on separate pages and separate approvals.

And the political failure, which is the most common. A federation over-devolves because devolution is popular and the cost of it is invisible — it shows up as under-provision spread thinly across everybody, never as a line somebody has to defend. The register is the countermeasure, and it works precisely because it makes the invisible cost a number on a page.

---

DELIGHT

What it feels like

There is a specific relief in the meeting where somebody says this should be central and somebody else says this should be local, and a third person asks what fraction of the benefit leaves the unit. The room changes temperature. Two positions that were about identity become two estimates about a function, and estimates can be wrong in public without anybody losing face.

Then the better pleasure, which arrives about a year later: the argument does not recur. It has a row, the row has a date, and everybody knows when it will be looked at again. A surprising amount of organisational unhappiness is not disagreement at all — it is the same disagreement arriving unscheduled, over and over, with no way to put it down.

And the quiet one. A member who was struggling gets lifted to the floor, and nobody has to describe them as struggling. There is no case to make, no plea, no gratitude owed. The index moved and the formula did the rest. The best thing about an equalisation that is arithmetic is that it lets a member be weak this year without being a supplicant, and any federation that has watched the other version happen knows exactly what that is worth.

---

OPERATIONALIZE THIS

At the level of finance

The instrument: a capability floor facility, governed by a published assignment register.

Two artifacts, one of which is a financial instrument. The register is free and does the analytical work. The facility is the money and makes the register honest, because a register that assigns a function downward to a member who cannot fund it has assigned it nowhere.

The structure. A standing facility inside the federation, funded by a levy on members whose capability index exceeds 100, paid out as an annual top-up to every member below a published floor F. Not a loan; not a grant; a formula.

The worked case, so the mechanics are visible. A federation of forty member societies, combined turnover 600.0 million pounds. One function under assignment. Capability indices run from 62 to 145 against a mean of 101.95 — a raw spread of 2.34. The function's annual surplus at stake is 225,000 pounds per member, 9,000,000 across the federation, and providing it efficiently costs 120,000 pounds per member per year.

The number that decides it, and it is one line:

      F*  =  100 x ( 1 − C / 2W* )

  C   the annual cost of providing the function efficiently, per member
  W*  the annual surplus at stake in providing it, per member

Raise the floor while the last index point of top-up removes more expected loss than it costs. Marginal loss removed per index point is 2W(1 − i/100)/100; marginal cost is C/100; they cross at F. For this federation, C / 2W = 0.2667, so F = 73.33.

That answer is lower than anybody's instinct and it is worth sitting with:

At the computed floor of 73.33 the facility costs 32,000 pounds a year, removes 36,865 pounds of expected loss, nets 4,865, and returns 1.15 pounds for every pound of top-up. Full levelling to 100 costs 385,200, removes 176,692, and returns 0.46 pounds per pound — it destroys 208,508 pounds of value a year while looking like the generous answer. The difference between the two policies is 213,372 pounds a year, and the generous-looking one is worse for every member including the weakest, because the money burned is money the federation cannot spend on the function itself.

The levy. Twenty members sit above index 100, holding 399 index points between them. At 80 pounds per index point the facility funds itself exactly. The strongest member pays 3,609 pounds a year — 1.60 percent of its own stake in the function, and 0.0053 percent of combined turnover across the whole federation. A federation that cannot find five thousandths of one percent of turnover to keep its weakest quarter capable has a different problem than this facility solves.

Balance-sheet treatment. The levy is an operating expense of the paying member, not a distribution — it purchases a capability the payer depends on, and the dependency is documentable through the register. The top-up is restricted income at the receiving member, recognised in the period, and it should be disclosed by name and amount. Where the function is an asset — a shared system, a licence, a certification — capitalise at the federation and license it down; do not let forty members capitalise forty copies.

The counterparty. The federation itself, or its shared-services entity. Never a bilateral transfer between two members: bilateral transfers create obligations and obligations create factions, which is precisely the failure a formula is there to prevent.

The first ninety days.

The one thing that must not be skipped. Publish the index before you publish the floor. A federation that sees the floor first reads the index as a judgement about its members. A federation that sees the index first reads the floor as the obvious response to something it has already looked at together, and that is the entire difference between a facility that is adopted and one that is voted down.

---

APPRECIATIVE QUESTIONS

Twelve, for a room

Discovery — what is already working

1. Name a function we hold at what turned out to be exactly the right level. How did it get there, and who noticed it was right?
2. Where have two of our units quietly cooperated on something without being told to — and what made that pairing work?
3. Which decision that we make centrally would we make exactly the same way if it were made locally? What does that tell us about that row?

Dream — what becomes possible

1. If every function we perform had a named level, two numbers, and a review date, what argument would we stop having?
2. Imagine our weakest member is as capable as our median member three years from now. What did we fund to make that true, and what did it cost?
3. If our members were free to form a boundary around a single shared function without asking anyone, what is the first thing two of them would share?

Design — what we build

1. Which one function are we most confident is at the wrong level, and what would we have to measure to be sure?
2. What is the best available proxy for how much of a benefit leaves the unit — something we already collect and could read this week?
3. Who should own the register, and what makes that person the right owner rather than the obvious one?

Destiny — how it holds

1. What has changed in the last three years about what our smallest member is capable of, and which rows should have moved because of it?
2. If the floor became something members lobbied for, what would we have seen six months earlier?
3. When we next disagree about centralisation, what will we point at — and who in this room would be able to point at it without us?

---

WORKS CITED

Pius XI (1931). Quadragesimo Anno, §79. The original formulation: it is an injustice to assign to a greater association what lesser and subordinate organisations can do.

Tiebout, C. M. (1956). "A Pure Theory of Local Expenditures." Journal of Political Economy, 64(5), 416–424.

Musgrave, R. A. (1959). The Theory of Public Finance. McGraw-Hill.

Ostrom, V., Tiebout, C. M. and Warren, R. (1961). "The Organization of Government in Metropolitan Areas: A Theoretical Inquiry." American Political Science Review, 55(4), 831–842.

Oates, W. E. (1969). "The Effects of Property Taxes and Local Public Spending on Property Values: An Empirical Study of Tax Capitalization and the Tiebout Hypothesis." Journal of Political Economy, 77(6), 957–971.

Olson, M. (1969). "The Principle of 'Fiscal Equivalence': The Division of Responsibilities among Different Levels of Government." American Economic Review, 59(2), 479–487.

Oates, W. E. (1972). Fiscal Federalism. Harcourt Brace Jovanovich.

Hamilton, B. W. (1975). "Zoning and Property Taxation in a System of Local Governments." Urban Studies, 12(2), 205–211.

Epple, D. and Zelenitz, A. (1981). "The Implications of Competition among Jurisdictions: Does Tiebout Need Politics?" Journal of Political Economy, 89(6), 1197–1217.

Ostrom, E. (1990). Governing the Commons: The Evolution of Institutions for Collective Action. Cambridge University Press. Design principle 8, nested enterprises.

Weingast, B. R. (1995). "The Economic Role of Political Institutions: Market-Preserving Federalism and Economic Development." Journal of Law, Economics, and Organization, 11(1), 1–31.

Alesina, A. and Spolaore, E. (1997). "On the Number and Size of Nations." Quarterly Journal of Economics, 112(4), 1027–1056.

Føllesdal, A. (1998). "Survey Article: Subsidiarity." Journal of Political Philosophy, 6(2), 190–218.

Oates, W. E. (1999). "An Essay on Fiscal Federalism." Journal of Economic Literature, 37(3), 1120–1149.

Frey, B. S. and Eichenberger, R. (1999). The New Democratic Federalism for Europe: Functional, Overlapping and Competing Jurisdictions. Edward Elgar.

Bardhan, P. (2002). "Decentralization of Governance and Development." Journal of Economic Perspectives, 16(4), 185–205.

Besley, T. and Coate, S. (2003). "Centralized versus Decentralized Provision of Local Public Goods: A Political Economy Approach." Journal of Public Economics, 87(12), 2611–2637.

Rhode, P. W. and Strumpf, K. S. (2003). "Assessing the Importance of Tiebout Sorting: Local Heterogeneity from 1850 to 1990." American Economic Review, 93(5), 1648–1677.

Oates, W. E. (2005). "Toward a Second-Generation Theory of Fiscal Federalism." International Tax and Public Finance, 12, 349–373.

Cooper, I. (2006). "The Watchdogs of Subsidiarity: National Parliaments and the Logic of Arguing in the EU." Journal of Common Market Studies, 44(2), 281–304.

Ostrom, E. (2010). "Beyond Markets and States: Polycentric Governance of Complex Economic Systems." American Economic Review, 100(3), 641–672.

Faguet, J.-P. (2012). Decentralization and Popular Democracy: Governance from Below in Bolivia. University of Michigan Press.

Treaty on European Union, Article 5(3); Protocol No. 2 on the Application of the Principles of Subsidiarity and Proportionality, Articles 6 and 7.

European Commission. Annual Report on the Application of the Principles of Subsidiarity and Proportionality. Successive editions, 2010 onward.

Federal Constitution of the Swiss Confederation, Articles 3, 5a, 43a and 135.

Swiss Federal Finance Administration. Finanzausgleich: Grundlagen und Kennzahlen. Annual resource-index and equalisation figures.

Swiss Federal Statistical Office. Amtliches Gemeindeverzeichnis der Schweiz. Annual commune counts.

KPMG. Clarity on Swiss Taxes. Annual survey of cantonal corporate income tax rates.

Basic Law for the Federal Republic of Germany, Article 107; Gesetz über den Finanzausgleich zwischen Bund und Ländern, as amended with effect from 2020.

German Federal Ministry of Finance. Vorläufige Abrechnung des Länderfinanzausgleichs. Annual.

Rabobank. Annual Report 2016. The merger of the local member banks and the central organisation into one legal entity, effective 1 January 2016.

Bundesverband der Deutschen Volksbanken und Raiffeisenbanken (BVR). Annual member statistics and the Sicherungseinrichtung, in operation since 1934.

United States Census Bureau. Geographical Mobility/Migration, CPS Annual Social and Economic Supplement; American Community Survey, commuting and state-of-birth tables; 2022 Census of Governments.

Note on figures. Every figure above is computed in lib/verify/VI_08.py and reproducible with python3 lib/verify.py VI.08. Inputs are printed there in three kinds, kept apart: legal constants, published series, and stated assumptions a reader is invited to change. The assignment rule's curvature term ρ is set to 1 by default and every threshold in the chapter moves if you move it. Chapter I.10 computes the coordination side of federation and is not repeated here.