Haute Lumière

Commerce · III.04 · MMXXVI · daylight

La Bourse  /  Volume III  /  Nº III.04  /  Ten concept briefs

TEN CONCEPT BRIEFS · Chapter III.04 — Credit as a Commons

One page each. A reader who reads only these ten pages has the chapter.

---

BRIEF 1 — Credit as a Commons

The idea. Most of the credit in a working economy is not made by banks. It is extended sideways between firms, as trade payables, and it behaves like a shared resource rather than a product.

A commons is a stock that many people draw on and that regenerates at a rate. A pasture regenerates through grass. A credit commons regenerates through trade between its members — every sale inside the circle is what allows an outstanding obligation to be worked off. That gives it a carrying capacity, and the capacity is not set by goodwill. It is set by how much the members actually buy from one another.

Worked example. A circle of 220 firms turns over €4,400,000 a year inside itself — €20,000 per member — and carries €640,000 of outstanding net debits. That is 14.55 percent of turnover, or 53.1 days. The debits are clearable because the trade exists to clear them. Double the debits without doubling the trade and they are no longer clearable at all.

Why it matters. It moves the question from is this credit good to can this population absorb this much credit, which is the question a regulator asks about a bank and almost nobody asks about a trading community.

You already know this because you have watched somebody's tab at a local place go quietly from he'll settle it to he's never going to settle it, and you knew which side of the line it was on long before anyone said so out loud.

---

BRIEF 2 — Multilateral Netting

The idea. When many parties owe many parties, the cash needed to settle all of it is far smaller than the sum of what is owed — and the way to find out how much smaller is to compute each party's net position.

Bilateral netting cancels what a pair owes each other. It is the obvious move and it usually does very little, because your supplier is rarely your customer. Multilateral netting computes, for each party, receivables minus payables. The positions sum to zero. The cash the whole system needs is the sum of the positive positions.

Worked example. Eight firms, sixteen obligations, €339,000 gross. Bilateral netting removes nothing at all — no pair owes in both directions. The multilateral net positions run from −€21,000 (the tannery) to +€28,000 (the maker) and sum to zero. Total cash required: €29,500.

  gross                €339,000
  cash to settle        €29,500
  released             €309,500
  each euro of cash clears  11.49  of trade

Why it matters. Whole chains of firms borrow from the same bank to pay one another. Netting first tells you exactly how much of that borrowing was never necessary.

You already know this because you have been in a group where four people were splitting bills and one person paid, and instead of four transfers there was one.

---

BRIEF 3 — Obligation Clearing

The idea. A loop of debt can be cancelled with no money at all.

If A owes B, B owes C, C owes D, D owes E and E owes A, then subtract the smallest amount on the loop from every edge. Each firm has been relieved of the same sum owed and owing, so nobody's net position changes by a single unit — and the gross obligations fall by that amount multiplied by the length of the loop. Nothing was paid. Nothing was borrowed.

Worked example. In the circle of eight, the loop tannery → mill → cutter → maker → shipper → tannery has a binding edge of €18,000. Cancelling it extinguishes €90,000, which is 26.55 percent of gross obligations, with no cash. Cancelling cycles greedily through the whole circle removes €272,000 — 80.24 percent — leaving an acyclic residue of €67,000.

The honest caveat. Greedy cancellation depends on the order cycles are found in. A different search order removes a different total. That is a property of the algorithm and it should be stated every time the figure is quoted.

Why it matters. Slovenia has run this as a national public service since 1991. It is not a proposal; it is infrastructure that most countries simply do not have.

You already know this because you have been owed by somebody you owed, and the two of you agreed to forget it — which is the same move with two parties instead of five.

---

BRIEF 4 — Netting Efficiency Is a Property of the Graph

The idea. How much a clearing system removes is decided by how densely the obligations loop — who trades with whom. It is never decided by the software.

  netting efficiency  =  1  −  (cash required to settle / gross obligations)

The numbers, side by side.

The spread between a national set-off and CLS is 86.0 points, and the arithmetic is identical in all five rows. CLS clears 96.0 percent of USD 8.0 trillion a day because its members are all dealers in the same instruments — maximum loop density. A national population of every firm in an economy is sparse, so it clears 10 to 15 percent.

Why it matters. Any promoter who quotes an efficiency without quoting a population is selling you a ledger. Ask for the graph.

You already know this because a lift-sharing scheme works brilliantly in one office and not at all in another, and the software was the same.

---

BRIEF 5 — The Clearing Coverage Ratio

The idea. A member's debit is only credit if the circle can clear it through trade. So measure it in months of that member's own in-network sales.

  clearing coverage ratio  =  debit balance / monthly in-network receipts

Worked example. The tannery holds a €21,000 debit against €49,000 of annual in-network receipts, which is €4,083.33 a month. Its coverage ratio is 5.14 months. Against a policy cap of 3.0 months — a limit of €12,250.00 — it is €8,750.00 over, by 71.4 percent.

Where the line sits. Under roughly three months, the balance is trade credit the circle works off in the ordinary course of business. Past it, the balance has become a loan: it will not be cleared by trading, and the circle has no machinery to call a loan. Nothing dramatic happens at the boundary, which is exactly why it has to be measured and published.

Why it matters. This single ratio is the carrying capacity of the commons, stated per member. It is the number a circle's monthly page should lead with, and it is cheap to compute from data the ledger already holds.

You already know this because you have judged whether somebody could repay by asking what they earn, not by asking what they owe.

---

BRIEF 6 — The Leverage Identity

The idea. A mutual credit circle is a bank read backwards — and by default it has no capital at all.

In a mutual ledger the balances sum to zero. So the outstanding credits equal the outstanding debits, exactly. The positive-balance holders are the depositors, the negative-balance holders are the borrowers, and between them sits an institution holding nothing against the book.

Worked example.

The repair, and it is small. A levy of 1.0 percent on in-network turnover raises €44,000 a year on €4,400,000 — a reserve ratio of 6.88 percent against the debits. It clears the Basel leverage floor equivalent in 0.44 years and a 13.0 percent CET1-equivalent in 1.89 years.

Why it matters. It explains why the right literature for mutual credit is Ostrom rather than Basel: with no capital, governance is the loss absorber. And it explains why almost every scheme should charge a levy it currently does not.

You already know this because you know what an uninsured warehouse is, and it is not a cheaper warehouse.

---

BRIEF 7 — Ostrom's Eight, Applied to Credit

The idea. Elinor Ostrom's contribution was never the claim that commons work. It was eight design principles specifying when they collapse. Credit has the same shape.

1. Clearly defined boundaries — the common bond; who may hold a debit.
2. Congruence with local conditions — debit lines set by in-network sales, so a member selling €4,083.33 a month does not get the line of one selling ten times as much.
3. Collective-choice arrangements — one member, one vote on the limits. Not one euro, one vote.
4. Monitoring accountable to members — balances and coverage ratios published to members, not only to staff.
5. Graduated sanctions — warning, frozen line, suspension, expulsion. Never straight to the extreme and never nothing.
6. Cheap, local conflict resolution — a members' panel in a fortnight, free.
7. Recognition of the right to organise — a legal form that holds, and a written tax treatment before the first trade.
8. Nested enterprises — circle inside federation inside reserve inside apex.

The worked failure. Timothy Guinnane's study of Raiffeisen's Irish transplant, 1894–1914, found societies breaking principles 1, 4 and 5 at once: lending outside the parish, monitoring nobody, sanctioning nothing. They did not survive. The German originals, on the same rules, did.

The ninth term credit needs. A pasture cannot be levered; a credit commons can. So the constitution has to name the levy, its percentage and the reserve ratio it targets. Ostrom's principles were written for stocks that cannot be multiplied by a ledger entry.

You already know this because every shared house you have lived in that worked had a rota, a way to complain, and a consequence — and every one that failed was missing one of the three.

---

BRIEF 8 — Joint Liability, and What It Actually Bought

The idea. Making borrowers liable for each other was the most famous innovation in modern credit. When it was finally tested experimentally, it did not pay for itself.

The founding logic. Raiffeisen's societies carried unlimited joint liability inside one parish. It worked — but Guinnane's evidence is that the mechanism was information, not the threat: the society knew things about a borrower no outside lender could learn at any price.

The experiment. Xavier Giné and Dean Karlan randomly removed group liability from existing Philippine lending groups and randomly assigned it in new areas. Removing it produced no increase in default over three years — and larger groups. In the expansion areas, group liability produced no default advantage and fewer new groups formed. It deterred clients and delivered nothing.

What did work. Benjamin Feigenberg, Erica Field and Rohini Pande randomised meeting frequency. Clients assigned to weekly rather than monthly meetings were about 3.0 times less likely to default on the following loan. The mechanism was repetition and the social interaction it produced — not the threat.

The measured cost. Fieldwork by Richard Montgomery and by Aminur Rahman documented peer seizure of household assets, coerced repayment and household conflict inside joint-liability groups. These costs are real, they fall on borrowers, and they are invisible in a lender's portfolio quality report.

You already know this because you have been in a team where one person's missed deadline was everyone's problem, and you know whether that made the team better or simply made it quieter.

---

BRIEF 9 — The Cooperative Loss Profile Is a Tail, Not a Mean

The idea. Cooperative lenders are not permanently safer than commercial ones. They are steadier, and steadiness is paid for in the calm years.

The numbers.

In a benign year the cooperative form loses 0.12 points more — a factor of 1.176. In the crisis year it loses 1.46 points less — a factor of 2.21.

The breakeven.

  crisis frequency at which the two cancel
     =  0.12 / (1.46 + 0.12)  =  7.59 %   — one year in 13.17

Over a decade containing two crisis years, the cooperative form is 1.96 points ahead cumulatively. Over a decade containing none, it loses.

Why it matters. It gives cooperative institutions an honest sentence to say about themselves, and an honest sentence survives the first quiet decade. We are not cheaper. We are steadier, and here is the crisis frequency at which that trade is worth making.

You already know this because you have paid an insurance premium for years without a claim and did not conclude the insurance was a bad product.

---

BRIEF 10 — A Repayment Rate Measures Collection, Not Return

The idea. How much of a loan book comes back tells you about the lender's asset quality. It tells you nothing about whether the borrower gained.

The two facts that look contradictory. Grameen-style lending reports repayment around 97.0 percent — an implied loss rate of 3.0 percent. And six randomised evaluations published together in January 2015, across Bosnia, Ethiopia, India, Mexico, Mongolia and Morocco, found modest business investment and no transformative effect on consumption, income, health, schooling or empowerment. Rachael Meager's pooled analysis of seven of those experiments puts the average consumption effect near zero and estimates it precisely.

The arithmetic that resolves them. In Hyderabad the take-up gap between treated and control areas was 8.8 percentage points, so scaling an intention-to-treat effect to those actually treated multiplies it by 11.36 — and an effect indistinguishable from zero, multiplied by eleven, is still indistinguishable from zero.

Then the price. Suresh de Mel, David McKenzie and Christopher Woodruff measured real returns to capital in Sri Lankan microenterprises at 4.6 to 5.3 percent a month — 55.2 to 63.6 percent a year on their own annualisation, 71.5 to 85.8 percent compounded — and found no positive return in enterprises owned by women. Against a 30.0 percent midpoint APR, the male-owned enterprises clear by 25.2 points and the female-owned ones are 30.0 points under water.

The denominator, stated. That is Sri Lanka, not Bangladesh, and a different population from Grameen's borrowers. It illustrates a mechanism rather than measuring a bank: a loan priced above the borrower's return is repaid out of something other than the investment.

Why it matters. Yunus's original observation at Jobra — 42 people held by USD 27.00 of debt, USD 0.6429 each — was right, and remains right. The binding constraint was the shape of the available capital. What did not hold was the claim that fixing the shape transforms the outcome on average.

You already know this because you have repaid something on time by going without, and nobody who saw the payment arrive learned anything about what it cost you.

---

All figures in these briefs are computed in lib/verify/III_04.py, with their inputs, units and sources printed first, and sourced in the chapter's Works Cited.