Haute Lumière

Commerce · VI.08 · MMXXVI · daylight

La Bourse  /  Volume VI  /  Nº VI.08  /  Workbook — the executive

WORKBOOK — THE CORPORATE EXECUTIVE

Chapter VI.08 · Federation and Subsidiarity

For the person who has to settle where a function sits, in a firm with a P&L, a board, and people on both sides of the argument who are not wrong. Ninety days from a standing dispute to a signed register.

---

THE PREMISE, STATED COMMERCIALLY

Your organisation is spending real money on a question it has never framed correctly. The question is not how centralised should we be. It is where does each function belong, and it has a different answer for each of the forty or so functions you actually perform.

The cost of getting it wrong does not appear as a line. That is the whole problem. Over-centralise and you pay in uniformity — a standard that fits the largest unit and nobody else, and the cost shows up as underperformance in the units that were forced into it. Over-devolve and you pay in under-provision — every unit rationally under-invests in anything whose benefit leaks, and the cost shows up nowhere at all. Neither has an owner, a report or a budget line, which is why the argument is settled by seniority and re-opened annually.

This workbook gives you two estimates per function, one inequality, a register, and a small financial instrument that makes the register honest. Ninety days. The first sixty cost nothing but attention.

---

PART ONE — DISCOVERY

Days 1–15: the list, and what is already right

Exercise 1.1 — Write the function list (one working session, four hours)

Not an org chart. A list of functions: things the organisation does, one per line. Purchasing. Brand. Pricing. Hiring. Safety standards. Regulatory compliance. Product specification. Customer data. Treasury. Training. Facilities. Supplier qualification. Incident response. Capital allocation. Twenty to sixty lines. If a line names a department rather than a decision, rewrite it.

Run this with the people who disagree with each other in the room. The list alone resolves a surprising share of the disagreement, because most centralisation arguments are two people holding different functions and using the same word.

Exercise 1.2 — Mark the settled rows (one hour)

Which rows would nobody move? Mark them, and for three of them write why you know. You are calibrating on cases where you already have the answer.

Exercise 1.3 — The four federations (reading, two hours)

Read how four real federations have answered this, because all four have paid for their answers.

• Switzerland. Articles 3, 5a and 43a of the constitution reserve to the cantons everything not assigned upward, and the country pays a visible price for it: effective combined corporate income tax 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. There it is the reservation of powers working.
• The European Union. Subsidiarity written into Article 5(3) with an actual procedure attached. Read the record in Part Two.
• Germany. About 18.5 billion euros redistributed between the Länder in 2023, Bavaria paying about 9.0 billion — 48.6 percent of the whole — and Berlin receiving about 3.6 billion. About 672 euros a head paid, 947 a head received, a ratio of 1.41. Published, arguable, litigated, and durable.
• The German cooperative banks. Everything local except one thing: the risk pool, assigned upward and held there since 1934. Member insolvencies costing a depositor money in the ninety years to 2024: zero.

The lesson to carry into your own list is the same in all four: none of them decided how centralised to be. They placed functions one at a time.

---

PART TWO — THE ARITHMETIC

Days 16–45: two estimates per row

Exercise 2.1 — Estimate s for every row (two sessions)

s is the fraction of a function's benefit that lands outside the unit that decides it. Assign each row to somebody who knows it, give them ten minutes, and require a route as well as a number.

• The leakage question. If this unit does this well, what share of the benefit goes to somebody who did not pay for it? Count sister units, future members, and the shared brand.
• The migration route. For anything invested in people, s is roughly the fraction who leave over the payback horizon. Payroll answers it in an hour. This is exactly how the public sector arrived at education: about 57 percent of Americans live in the state they were born in, so s for schooling is about 0.43, and every federation on earth has moved education finance upward.
• The complaint route. Of those who complain when this is done badly, what share are outside the unit?

Exercise 2.2 — Estimate h for every row (one session)

h is the coefficient of variation of the level each unit would choose alone. Where the function is already local, measure it from what units already do — that is h revealed rather than stated, and it is better. Six branches choosing weekly cleaning frequencies of 2, 3, 3, 4, 5 and 7 give a mean of 4.00, a sample standard deviation of 1.79, and h = 0.45; one uniform frequency costs 0.2000 of that function's value.

Exercise 2.3 — Run the rule (two hours)

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

  they are equal at   s* = h / (1 + h)

The value at stake cancels, so you can apply this to functions nobody has costed — which is most of them.

The threshold table, and it will surprise your leadership team:

  h      s*
  0.10   9.1%
  0.20   16.7%
  0.30   23.1%
  0.40   28.6%
  0.50   33.3%

Worked, on a compliance function: s = 0.35, h = 0.28. Shortfall 0.35/0.65 = 0.5385. Assigning down costs 0.2899; assigning up costs 0.0784; the threshold is 0.28/1.28 = 0.2188. The ratio is 3.70, and that is the sentence for the paper: leaving this with the unit costs us 3.70 times what standardising it costs, and the difference is under-provision nobody currently reports.

Exercise 2.4 — Mark the contested rows and leave them (one hour)

Where s exceeds s* by less than a third, mark the row contested and do not move it. A leadership team that moves its near-misses spends its whole political capital on the rows where the answer barely matters.

Exercise 2.5 — Kill two arguments you were going to have (one hour)

Two arguments are now over and you should retire them explicitly.

"People will vote with their feet." Tiebout sorting is real and slow. In the United States 7.8 percent of the population moved in 2023–24, about 64 percent of movers stayed inside the same county, so 7.8 percent times 36.0 percent = 2.808 percent crossed a jurisdiction — a half-life of 24.3 years. In 1948, at 20.2 percent movers, it was 9.2 years; the mechanism is 2.65 times slower than it was. It will not correct anything inside your planning horizon.

"Subsidiarity is a principle, not a calculation." The European Union's early warning system is the most serious attempt anyone has made to run subsidiarity as a principle with a procedure. Sixteen years, three yellow cards, mean margin 2.67 votes — less than two national parliaments — one card every 5.33 years, zero orange cards against a threshold of 29, one proposal withdrawn out of three at 33.3 percent, and zero acts annulled by the Court. A principle with a procedure and no arithmetic behind it fires once every five years and changes one proposal in three.

---

PART THREE — DESIGN

Days 46–75: the register and the instrument

Exercise 3.1 — Build the register (one session)

One page, one row per function: level, s, h, s*, verdict, contested flag, date, and the name of whoever estimated. The name is load-bearing. An estimate with nobody's name on it cannot be challenged, so it hardens into a fact.

Exercise 3.2 — Separate assignment from execution (one hour)

The rule says where 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. Conflating the two is how a correct assignment becomes an unpopular head office, and it is the single most common way this work is wasted.

Go through your proposed moves and write, for each, the decision level and the execution level separately. Half of them will turn out not to need anybody to move desks.

Exercise 3.3 — Build the capability index (one session)

One index, across units, for the capacity to fund and perform the functions you are leaving local. Revenue per unit, contribution, headcount, qualified staff — whatever is honest and already collected. Rebase to a mean of 100. Publish it unranked, as a distribution with each unit told its own number.

Then look at the spread, because this is the part of subsidiarity nobody says out loud. A function assigned to the lowest capable level goes, by construction, to whoever is already strong. Switzerland's resource index runs above 250 at the top and near 67 at the bottom — a raw spread of 3.73 to one — and a unit funding 0.67 of the efficient level loses (1 − 0.67)² = 10.89 percent of that function's value, in the place least able to bear it.

Exercise 3.4 — Compute the floor (two hours)

Not full equalisation. A floor, at a computed height.

      F*  =  100 x ( 1 − C / 2W* )

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

Worked on a forty-member federation with indices from 62 to 145, W of 225,000 pounds per member and C of 120,000: C / 2W = 0.2667, so F* = 73.33.

At 73.33 the facility costs 32,000 pounds a year, removes 36,865 of expected loss, nets 4,865, and returns 1.15 pounds per pound. Full levelling to 100 returns 0.46 pounds per pound and destroys 208,508 a year — a difference of 213,372 — while looking like the generous answer.

And the case the formula also settles: if W were only 70,000, then C / 2W = 0.857 and F* = 14.3, below every member's index. The correct floor would be no floor at all, and the right conclusion would be that the function costs nearly what it is worth and should be shared, re-scoped or dropped. A formula that tells you when not to build the facility is worth more than one that only tells you how big to make it.

Exercise 3.5 — Set the levy (one hour)

In the worked case, 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 federation.

Put that last figure in the paper. A group that cannot find five thousandths of one percent of turnover to keep its weakest quarter capable has a different problem than this facility solves.

---

PART FOUR — DESTINY

Days 76–90: sign it, and make it survive you

Exercise 4.1 — Publish the index before the floor (sequencing, not effort)

This is the one thing that must not be reordered. A group that sees the floor first reads the index as a judgement about its members. A group that sees the index first reads the floor as the obvious response to something everyone has already looked at together. That sequence is the difference between a facility that is adopted and one that is voted down.

Exercise 4.2 — Move two rows, not ten (one session)

Move the two clearest. Leave the contested ones where they are with a review date. Ten moves at once produces a reorganisation, and a reorganisation buries the arithmetic that justified it.

Exercise 4.3 — Set the review on a date (ten minutes)

Annual, in the calendar, short. Anything reviewed by exception is reviewed when somebody is angry, and angry reviews move functions for reasons the rule cannot see.

---

THE FAILURE MODES, NAMED

Capability moves and the register does not follow. The big one, and it runs in the direction nobody expects. Rabobank's local member banks fell from 174 in 2010 to 106 in 2015 — 39.1 percent in five years — and in 2016 the remaining banks merged into one legal entity, because supervision and capital rules had raised the cost of being capable. The German cooperative banks went from about 7,100 around 1970 to 672 at the end of 2024: a fall of 90.5 percent over 54 years, compounding at 4.27 percent a year. The lowest capable level rises under you while your structure stays where it was.

And the hopeful half of the same record, which belongs in the same breath: through a 90.5 percent consolidation of the members, the functions assigned to the federation did not move, because they had been assigned about the function rather than about the members.

The centre learns to differentiate and nobody re-reads the register. Oates's theorem needs a centre that must give one answer. Software has made differentiation nearly free. Several rows should move up the year you acquire that capability, and almost nobody looks.

The floor becomes a grant. A floor tied to a published index is arithmetic. A top-up made "just this year" for a unit in trouble is a grant, and grants are permanent. Separate pages, separate approvals.

The register is written and never used. The countermeasure is that the two estimates go into the standing template for any proposal to move a function. Anything in the template persists; anything reviewed by exception does not.

---

THE NINETY DAYS ON ONE PAGE

---

BOARD PAPER TEMPLATE

One page. Six headings, in this order.

1. What we are deciding. One function, named. Not a philosophy.
2. The two estimates. s, h, the route for each, the name against each.
3. The arithmetic. s, both losses, and the ratio between them. One sentence: leaving this local costs us 3.70 times what standardising it costs.*
4. Decision and execution, separately. Where it is decided; where it is done.
5. The capability position. The index spread, the computed F*, the annual cost as a share of turnover.
6. The review date. When this row is looked at again, and what would move it.

---

APPRECIATIVE QUESTIONS FOR YOUR LEADERSHIP TEAM

1. Which function are we holding at exactly the right level, and how did it get there?
2. Where have two units quietly cooperated without being told to, and what made that work?
3. What argument would we stop having if every function had a level, two numbers and a date?
4. If our weakest unit were as capable as our median unit in three years, what would we attempt that we currently do not?
5. What can our smallest unit do now that it could not do three years ago, and which row should move because of it?
6. When we next disagree about centralisation, what will we point at — and who here could point at it without the rest of us in the room?