The Layering of Credit: Why Trust Can Be Transferred but Never Outsourced

1. One Word, Two Ledgers

In English, credit is both a moral word and a financial one. It comes from the Latin credere — to believe. The same root gives us creed, credential, incredible. This is not an etymological curiosity. In the centuries before finance became a technology, the entire content of a loan was a single sentence: I believe you will repay.

That sentence was later split into two ledgers. The first is money, recorded on a balance sheet. The second is trust, recorded somewhere else — sometimes in a relationship, sometimes in a reputation, sometimes in an institution, but never on any sheet at all.

There is the difficulty. In five hundred years we invented double-entry bookkeeping, central banks, deposit insurance, credit ratings and cross-border settlement systems for the first ledger. For the second we invented far less, and most of what we invented merely re-recorded it elsewhere rather than recording it properly. The result is counter-intuitive: the more sophisticated the financial technology, the lower the trust cost of any single transaction — and the higher the system's total demand for trust.

In the autumn of 2008 the overnight lending market froze. The problem was not a shortage of money; a central bank can print money, and did. The problem was that banks did not know whether the bank across the street would still exist tomorrow. Money can be printed. Trust cannot.

This essay asks a narrower question. When we hand trust over to institutions, is trust replaced, or merely moved down a layer? And why, after the move, does it not become any cheaper?

2. Three Pillars of Personal Trust, and Why It Cannot Scale

Before institutions, trust ran on three things, and all three had to be present at once.

The first is repetition. Cheat me now and no one will trade with you later. Repetition converts a one-shot game into a long one, discounting the gains of deception against future losses.

The second is observability. Your conduct must be seen, and remembered. Reputation works in a village not because villagers are more virtuous but because information travels fast and accurately.

The third is community sanction. The cost of being found out must be real — not abstract disapproval but concrete exclusion: no grain to borrow, no family willing to marry in, no guild that will admit you.

All three pillars share one property: they are local. Widen the circle and verification costs rise exponentially. You can know who is reliable among thirty people; you cannot know among ten thousand. This is not a limit of information technology but a structural one — judging a person requires a continuous record of that person, and records accumulate only where relationships persist.

Adam Smith noticed a by-product of commercial society: frequent exchange turns reputation into an asset, because the expectation of meeting again is stretched out and cheating becomes unprofitable. This observation is often cited to show that markets make people honest. It carries a premise that is easily skipped: exchange must be repeated and observable. Smith was describing a small world of merchants, not an anonymous mass market.

Karl Polanyi put the matter more bluntly in The Great Transformation. He distinguished three forms of economic integration: reciprocity, redistribution, and market exchange. The first two are embedded in durable, predictable social relations — reciprocity presupposes that we will meet again. The market pulls exchange out of relationship, replaces relational signals with price signals, and raises efficiency sharply; what it pulls out along with them is the very carrier on which trust had depended.

So the modern predicament is not that people became untrustworthy. Human nature did not deteriorate in the eighteenth century. The predicament is that the radius of exchange exceeded the radius of trust. You must transact with someone you have never met, will never meet again, and cannot sanction. Not one of the three pillars is present.

3. The First Transfer: Handing Trust to Records

The earliest solution was not an institution. It was a piece of paper.

The bill of exchange is an ingenious device: payment is made in one place, and the bill is redeemable in another, by another party, at a later date. It did not eliminate risk. It exchanged the risk of "this person's character" for the risk of "whether this paper will be honoured", and then passed the obligation of honouring it to the next holder. Its return appeared as a difference in exchange rates rather than as interest — and in an era when the church prohibited usury, that formal distinction determined whether the instrument could legally exist.

This is the first layering of trust. You no longer need to believe the man in front of you; you need only believe the paper, and the third party willing to discount it. Trust is not cancelled. It is cut into smaller shares and distributed along the chain.

Then came three harder things.

Accounting. Double-entry bookkeeping compresses a person's dealings into a format a third party can read. Its significance is not accuracy of calculation but the fact that it makes "what happened" externally checkable. The object of trust shifts from the person to the record.

Registration. Land registries, company registers, public mortgage filings. All answer the same question: how does a stranger confirm that this thing is genuinely his? The answer is not to ask the man but to ask the book.

Enforcement. The hardest layer, because it does not require you to believe the other party at all — only to believe that breach will have consequences. Courts, arbitration and insolvency proceedings are, in essence, the replacement of trust with the predictable use of force.

Stack the three and modern finance becomes possible. But notice what each layer actually does: it pushes the question "whom do you trust" back by exactly one step.

You trust the bank because deposit insurance stands behind it; you trust deposit insurance because a central bank stands behind it; you trust the central bank because the state stands behind it; you trust the state — and here the chain must stop.

Where it stops is usually not a rational argument but a convention, or a coercion. This is not a defect that can be patched. It is the necessary terminus of any chain of trust. Wittgenstein made a similar point about rule-following: all justification comes to an end, and the end is not another reason but the fact that this is what we do.

4. The Case Against: Low Trust Is Often Rational, and Verification Is Structurally Flawed

A natural inference follows: if institutions can substitute for trust, a low-trust society need only build the institutions. That inference is thoroughly wrong, and it is worth taking apart piece by piece.

First objection: low trust is often an accurate reading of the environment, not a cultural deficiency.

Acemoglu and Robinson distinguish inclusive from extractive institutions. Under extractive institutions, whether a contract is enforced depends on the parties' relationship to power rather than on the contract itself. In such an environment, distrust is not ignorance but a high-information judgement: you have observed the actual rules and adjusted accordingly.

This overturns half of the "trust is a virtue" narrative. When a society distrusts strangers pervasively, the likelier explanation is not moral decline but the genuine absence of consequences for breach. Demanding that people become trusting first and build institutions second reverses the causal order. Trust is a product of institutions, not their precondition.

Second objection: verification is not free, and the verifier must itself be trusted.

This is the weakest joint in the whole outsourcing scheme. Before 2008, rating agencies were paid by the issuers whose securities they rated. Auditors are hired and paid by the companies they audit. This is not the moral corruption of a few individuals; it is a structural necessity. Whoever pays holds part of the verifier's attention.

The longer the chain of verification, therefore, the easier it is to rent one of its links. You can add regulators, but regulators must be overseen, and overseers must be funded. Each layer solves a little and adds a new node that must now be trusted.

Third objection: high trust has costs of its own, and accountability cannot be replaced by trust.

The costs of a high-trust society are routinely ignored. Closed networks are more prone to collusion, and fraud survives longer in an environment where everyone believes everyone — Madoff's scheme ran for two decades largely on the strength of his standing within a tight-knit community. For outsiders, a high-trust network is an invisible wall: it runs on long relationships, and newcomers have none.

Raising social trust is therefore not an unconditionally good goal. The real questions are whether that trust is open to those who exit, and whether it conceals conduct that ought to be answered for.

5. The Judgement: Three Tests and One Limit

Put together, the conclusion tightens into a single sentence: trust is not a moral resource but an infrastructure. It cannot be outsourced, only transferred down a layer; and every transfer lowers the cost of a single transaction while adding a node that must now be trusted.

Three tests follow, for judging whether a given arrangement of trust is sound.

1. Can breach be discovered cheaply? The question is not whether rules exist but what an ordinary person must pay to learn that the other side has broken them. The deeper the asymmetry of information, the more decorative the rules.
2. Does discovery carry a real consequence? The consequence must be concrete and enforceable, not nominal. A contract you can sue on but cannot execute is no contract.
3. Is exit possible? This is the most neglected test. If you cannot leave, your relation to the other party is not trust but dependence — and dependence does not require trust, only compliance.

The limit is equally clear. Institutional trust works best in transactions that are infrequent, large, and exit-able on both sides — cross-border trade, large credit facilities, real estate registries. In settings that are frequent, small, and expensive to exit, personal trust is cheaper. This explains why relationship-based exchange has never really disappeared from retail, services and informal lending: not because those participants are backward, but because in those settings the unit cost of institutions exceeds the unit cost of relationships.

Augustine Sedgewick's Coffeeland supplies the extreme counter-case. A coffee tree takes four to five years to bear, so the grower must invest years ahead of income — which means credit; credit requires collateral; collateral requires title; title requires registration; and registration requires excluding the people who had lived on that land for generations. El Salvador's coffee economy thus grew into a system of debt peonage, in which labour was repaid out of future work against debts recorded unilaterally by the employer. Every layer of "credit technology" was present — contract, ledger, registry, enforcement — and their combined effect was not to make trust unnecessary but to strip one party of any capacity to verify or to leave.

This is the coldest face of the word credit. When the producers of the record and the people recorded are not the same people, and those recorded cannot check the record, credit stops meaning "I believe you will repay" and starts meaning "you will repay, and the ledger is in my hand".

Return, finally, to the etymology. In the first book of the Treatise, Hume argues that causal inference rests not on reason but on custom: we expect the sun to rise tomorrow not because we have proved that it must, but because it always has — and that expectation itself admits of no rational justification. He called this habit "the great guide of life", a phrase that appears in his later rewriting of the argument, An Enquiry Concerning Human Understanding.

If even the sunrise rests on an unproven reliance, then demanding that institutions deliver certainty without trust is demanding something that does not exist. The real content of modernity is not that trust became unnecessary, but that its object shifted from persons to records — and the producers of records are still persons.

So the question was never how to make people more trustworthy. It is how to make the producers of records unable to revise them unilaterally, and whether those recorded have standing to object when the record is wrong. The first is a technical problem. The second is a political one. Confusing the two has been the most expensive misjudgement of the last five hundred years.