Bank-Based and Market-Based: Where Credit Is Judged

In a bank-based arrangement, most saving reaches borrowers across the balance sheets of institutions, and the judgement about a borrower is made inside one lender and stays there. In a market-based arrangement, most saving reaches borrowers as securities that many parties hold, and the judgement is made in public and struck again every time the security changes hands. Where that judgement sits decides how trouble becomes visible.

Neither arrangement is safer, cheaper, more modern or better run than the other. Quality is not what separates the two. The separation is the location of a single decision, and everything else falls out of that one placement.

What does every financial system have to decide, and who decides it?

Somebody, somewhere, has to work out whether a borrower is worth lending to. There is no arrangement of savers and borrowers that escapes this. Money moves from a household with more than it needs today to a business that needs more than it has, and between those two sits a question nobody can skip: will this come back?

Where that question gets answered is the one respect in which the two arrangements differ. In one, it is answered inside a single institution, which investigates the borrower, writes down what it found, and keeps the paper. In the other, it is answered by a crowd of parties who never meet the borrower and never meet each other, and their answer takes the form of a price. Six further differences follow, and not one of them is independent of the first.

Consider an ordinary street. The grocer who lets a household run a monthly tab has made a credit decision. He asked around once, watched for a year, and now the whole judgement lives in a notebook behind his counter. Two lanes away, a second-hand scooter is being sold in the open, and eight people are calling out what they will pay. Everything anybody knows about that scooter ends up compressed into the number it finally sells at, and every one of those eight hears it. Same decision, two completely different places for it to live.

What is a bank-based arrangement, stated completely enough to use?

In a bank-based arrangement, savings reach borrowers as claims on institutions. A household puts money into an institution, so the household holds a claim on the institution. The institution then lends that money out, so the institution holds the claim on the borrower. The saver and the borrower never face each other and, in the ordinary case, never learn of each other.

Claims on institutions produce a very particular way of working. The institution investigates the borrower once, properly, before lending. The institution decides once. The price is set once, in a conversation with the borrower. And then it keeps what it decided, usually for years. The advance is held to maturityKept on the books until the borrower repays, rather than sold on to somebody else partway through. The lender carries it at what it decided, not at what somebody would pay today. rather than passed along, and nobody stands over the institution each morning asking what the advance is worth today.

Everything distinctive about a bank-based arrangement follows from one property: what the lender learned about the borrower stays with the lender. The site visit, the awkward conversation about last year, the fact that the promoter paid late twice in 2019 and then never again, the way the stock moves before a festival season. None of it is published. None of it is checkable by anybody outside. The knowledge is not secret in any sinister sense. It belongs to the institution that went and found it out, and it is worth something precisely because nobody else has it.

What is a market-based arrangement, stated completely enough to use?

In a market-based arrangement, savings reach borrowers as securities. The borrower issues a claim on itself, and that claim is held by many parties at once, each of whom may sell it on to another party at any time. The saver holds the borrower's paper directly rather than holding a claim on an institution that holds the borrower's paper.

Securities held by many parties produce the mirror image of that working method. Nobody investigates once and keeps the result. Whoever is buying forms a view, whoever is selling forms a different one, and the transaction between them settles at a number. Then it happens again, and again, with different parties, and the number moves. Fresh parties striking a fresh number is repricingA claim being valued afresh by whoever happens to be buying and selling it at that moment, rather than carrying the value it was given when it was first created., and repricing is not an occasional event in a market-based arrangement. Repricing is the arrangement itself.

In a market-based arrangement, whatever anybody learns about the borrower shows up in a price that everybody can see. Everything distinctive about the arrangement follows from that mirror-image property. A party who works out that a borrower is in difficulty cannot keep the finding to itself, because the only way to act on the finding is to sell, and selling moves the price. The information gets out through the act of using it. Nobody publishes it and nobody has to.

Where does the judgement about a borrower sit, and who gets to see it?

With the two set side by side on this one question, the rest becomes predictable. In a bank-based arrangement, the judgement sits inside one lender, in a file nobody outside the lender reads. In a market-based arrangement, the judgement sits across many parties, in a price anybody can look up in a second.

For the borrower, that difference decides whether it explains itself once or forever. A borrower dealing with one institution explains itself deeply, in private, to a party that will keep the information and use it again next time. A borrower funded in a market explains itself continuously, in public, to parties it will never meet, most of whom will never speak to it and all of whom can walk away without a conversation. Neither is the easier life. One is a long interrogation gone through once; the other is a short interrogation that never stops.

Before its securities are listed, and for as long as they stay listed, a company must publish certain things. Those disclosureWhat a company issuing securities is required to publish, and to keep publishing, so that somebody with no relationship to it can form a view. The word covers the obligation, not any particular document or timetable. requirements are set by the Securities and Exchange Board of India (SEBI) at sebi.gov.in. The requirements exist precisely because the market-based arrangement asks strangers to form a view, and a stranger with nothing to read cannot form one.

How does the price of credit get found in each arrangement?

In a bank-based arrangement, the price of a particular credit is negotiated. Two parties sit down, one of them knows a great deal about the other, and they arrive at a number that reflects everything the lender has learned. The negotiated number is then known to exactly those two parties. Nobody else can check it, compare it, or use it as a reference for anything.

In a market-based arrangement, the price is discovered by parties bidding against one another, and the result is published to anybody who wants it. No individual party in that crowd knows anything like what the single lender knew. The crowd has numbers, arithmetic, and each other instead.

Each way of finding a price gives up exactly what the other one gains. A negotiated price can carry things a stranger could never find out, including the kind of judgement that never goes into a document at all. Somebody who knows nothing whatsoever about the borrower can check a published price. So the price travels. A party in another city can use it on a Tuesday afternoon with no phone call to anybody. Each is bought with the other, so neither is an advantage on its own.

How does a problem become visible, and how fast?

One distinction matters more than the rest, and it is the one most often read backwards. A traded claim reprices the moment enough parties change their minds about it, and every party watching sees the change the same day. A held loan shows nothing at all, for months, until the lender does one of three things: records that the advance has stopped performing, reports a figure, or sets an amount aside against it.

Each of the three acts has a name worth having. ClassificationThe point at which a lender formally records that an advance has stopped performing. Before that record is made, the lender's own books show the advance exactly as they did the month before. is the record that an advance has stopped performing. A provisionAn amount a lender sets against advances it expects will not be recovered in full. A provision reduces the reported value of the book without anything having been sold or written off. is an amount set against advances expected not to come back in full. And a reporting dateThe date as at which a published figure was struck. Every figure in a set of accounts describes a position on that one date, not the days before it or after it. is the date as at which a published figure was struck. The step is the distance between two readings taken a long way apart, so a held book that finally shows trouble shows it as a step.

The two arrangements do not differ in whether trouble happens. The difference is how quickly and how publicly trouble appears, and anybody who mistakes the second for the first has drawn a conclusion the evidence does not support. Nothing on the traded side is riskier for having been measured more often, and nothing on the held side is steadier for having been measured less. Hold that sentence; the failure block below is built entirely on it.

The everyday version is sitting in most households. The balance in a savings account updates every time it is checked. Gold kept at home does not change value on the day the rate is checked, but its value has been moving all along; there was simply no reading. Nobody would conclude that the gold is the steadier asset because the passbook moved more often, and yet that is precisely the conclusion people reach when they set a lender's book against a traded book.

What does each arrangement do when it comes under strain?

A lender holding a claim it never has to sell can wait. The lender can renegotiate, extend the schedule, accept less this year in return for more next year, and keep a borrower alive that would otherwise stop existing. Nobody is forcing a price on that lender today, so nobody is forcing a decision either.

Waiting is not one of the things a market does, so a market cannot wait. The moment enough parties change their minds, the price moves, and a borrower who needs to raise more finds the terms have changed underneath it. The same news that made the borrower need the money made the parties reluctant to give it, so the terms change fastest at exactly the moment the borrower most needs them.

Each arrangement's strength under strain is the very same property as its weakness, seen from the other side. The lender's freedom to wait is what keeps a good borrower alive through a bad year, and it is also what lets a problem sit unrecognised. The market's refusal to wait is what tells the truth sooner and to everybody at once, and it is also what shuts the door precisely when the door matters most. The two arrangements do not offer two lists of advantages and drawbacks to be totted up. There is one property on each side, and it faces both ways.

What does each arrangement need before it can work at all?

Neither arrangement is available for the asking. Each rests on conditions that took decades to build wherever they exist, and a place that lacks them cannot simply decide to have the arrangement anyway.

A market-based arrangement needs three things. The first is information a stranger can check for himself. A party in another city, forming a view on a borrower it will never visit, has nothing else to work with. The second is enough parties dealing for a price to mean anything. A price struck between two parties once a fortnight is a rumour rather than a price. The third is a way to enforce a claim held by somebody who has never met the borrower. A claim nobody can collect on is not a claim.

A bank-based arrangement needs a different three. The first is institutions carrying capital of their own. An institution that holds claims must be able to absorb the ones that go wrong. The second is a cheap and repeatable way of gathering savings. The whole arrangement runs on money the institution has to attract before it can lend. The third is the capacity to investigate borrowers over and over again. Investigation is expensive and slow, and it is the actual product being sold.

Neither set of conditions is free, and that is why arrangements differ between places rather than because somebody sat down and chose one. A place without checkable information about its companies and without enough parties dealing will run on institutions whether it admires markets or not. A place without institutions carrying capital will find borrowers going to whoever will hold the paper. The arrangement follows the conditions.

Do the two arrangements ever meet in the same place?

The two meet constantly, and the clearest place to watch it happen is inside a single balance sheet. Suvarna Commercial Bank Limited, an invented listed commercial bank used here for teaching, reports total assets of Rs 2,40,000 crore. Of that, Rs 1,44,000 crore is advances. The bank investigated those advances itself and holds them itself. Dividing gives Rs 60.00/- of advances for every Rs 100.00/- of assets.

The same bank reports Rs 60,000 crore of market instrumentsSecurities an institution holds on its own balance sheet, valued by whatever other parties are willing to pay for them rather than by anything the institution itself decided. on that same balance sheet. Dividing again: Rs 60,000 crore over Rs 2,40,000 crore is 25.0 per cent of total assets, or Rs 25.00/- for every Rs 100.00/- of assets. The remaining Rs 36,000 crore, or Rs 15.00/- in every hundred, is everything else the bank holds, and the underlying record does not break that remainder down.

One balance sheet carries both: a book whose condition appears only when the lender reports it, and a book whose value moves whether the lender likes it or not. Trouble therefore arrives at two different speeds inside a single institution. A reader who comes away thinking the two arrangements are separate worlds has been taught something false. Bank-based and market-based are two ways of holding a claim, and most large institutions hold claims both ways at once.

Now put the two books side by side and read what the lender publishes about each. On the advances, Suvarna Commercial Bank Limited reports gross non-performing advances of Rs 6,480 crore, or 4.50 per cent of gross advances of Rs 1,44,000 crore. Coverage is 70.0 per cent of those gross non-performing advances, so Rs 4,536 crore is held against them, leaving Rs 1,944 crore net. Net advances are Rs 1,44,000 crore less Rs 4,536 crore, or Rs 1,39,464 crore. Rs 1,944 crore over Rs 1,39,464 crore is 1.39 per cent on net advances. Notice that the denominator moved between those two ratios, and had to.

Every figure in the first eight rows of that table has one thing in common, and it is worth saying out loud: not one of them existed anywhere until the lender computed it and published it. The 4.50 per cent, the 70.0 per cent coverage, the Rs 1,944 crore and the 1.39 per cent were all struck inside the institution. Compare that with the last two rows. Kaveri Stock Exchange Limited, also invented, reports turnover of Rs 48,00,000 crore in the stated year, on which a transaction fee of 0.00325 per cent produces Rs 156 crore. The revenue is not the point. The point is that every rupee of that Rs 48,00,000 crore is a claim being priced again by somebody, and no institution had to sit down and decide any of it.

Which arrangement does a real economy run on?

Which arrangement any real economy actually runs on, and in what proportion, is a measured fact about a particular place as at a particular date. It moves. The proportion is measured in more than one way, and different measurements disagree in ways that matter. No figure of that kind should be recalled and written down.

A figure of that kind, recalled rather than looked up, would be wrong rather than merely old, and it would be wrong in a form that reads as authoritative. For the domestic series, the Reserve Bank of India publishes at rbi.org.in and maintains a database site at dbie.rbi.org.in. For a comparison drawn across countries, the International Monetary Fund publishes at imf.org.

When such a figure is found, two questions decide whether it is usable. The first is what exactly was counted. A measure that counts outstanding advances is not the same as one counting new issuance, and a measure counting listed securities is not the same as one counting all securities. The second is as at when. A proportion struck three years ago describes a place that no longer exists in quite that form. Both questions come before the number is used for anything.

What lies outside this comparison?