Credit Crunch: When Lending Stops Regardless of Rates

A credit crunch is lending falling because lenders pulled back, not because borrowers stopped asking. Which side stopped is the definition, and the published lending figure looks the same either way. Which side stopped is also why a crunch can run right through a period in which the price of credit is being cut, and why cutting it further does not end one.

Two ideas already established hold that answer up. The first is that a market has two sides, and an observed quantity is the outcome of both of them moving at once, so a quantity alone can never say which one did the moving. The second is the policy rate and the lag it works through. A crunch is the case where the lever is pulled properly and the price it acts on is not what stopped the lending. The whole subject is narrow: one number, two causes, and three observations that begin to separate them.

Which side stopped, and why is that the whole definition?

Lending in the Republic of Sankhya, an invented country used here for teaching, fell hard over a single year. The words describing the two possible events are not interchangeable, and the responses to them are not either. So before anything at all can be said about that fall, it must be established which of the two produced it.

The first event is that the lenders stepped back. The people who wanted to borrow were still there, still turning up, still asking for the same amounts as last year. The change sat on the other side of the desk: the lender became less willing to part with money, and started saying no to applications it would have said yes to twelve months earlier. A lender stepping back like that is a credit crunch, and the contraction is in the supply of credit.

The second event is that the borrowers stopped asking. The lender's willingness never moved. The lender would have approved exactly what it approved last year, on the same terms, to the same sort of applicant. But the applicants did not come. A shopkeeper who was going to add a second floor decided to wait. A workshop that borrowed every year for working capitalThe money a business needs to fund the gap between paying for stock, wages and materials and being paid by its own customers. Working capital is borrowed to run the business rather than to expand it. did not need as much because it was buying less stock. Borrowers going quiet like that is weak demand for credit, and nothing about the supply of credit has changed at all.

The fall in lending is measured on the same axis in both cases and comes out at exactly the same figure, so the lending figure by itself can never say which of the two is in view. This is not a subtle distinction that matters only to specialists. The distinction is the difference between a market where willing borrowers are being refused and a market where nobody is asking, and those are close to opposite situations wearing the same number.

Outside the language of economics, the same problem looks like this. A moneylender in a market town lends to stallholders, and last month she wrote fifteen loans and this month she wrote four. Two stories fit. In the first, a fire went through one row of stalls and she has decided she cannot judge who will still be trading in six months, so she is turning people away at the door. In the second, the wholesale market shut for renovation, so the stallholders have nothing extra to buy and eleven of them simply did not come and ask. Her ledger reads four either way. An observer standing outside the shop and counting loans cannot tell which of the two towns it is.

Why does a higher rate stop being able to fix it?

The instinct that a price clears any market is a good one and it runs out here, so it is worth being precise about where. A rate is the price of credit. Raising it pays a lender more for parting with money, and in the ordinary case that is enough: a lender who wanted a bit more to make the loan worthwhile gets a bit more and makes the loan.

Now change one thing. The lender is no longer haggling about how much it earns. The lender has become unsure whether the money comes back at all. Once that is the question, look at what a higher rate actually offers: a larger payment, in the event that payments are made. A higher rate is more of a payment the lender has stopped believing in. The doubt is not about the size of the return but about its existence. So adding to a sum that may be multiplied by nothing does not move the lender, and no rate on offer is high enough to change that.

A lender who doubts repayment does not want a higher price, it wants a different borrower, and there is no rate that turns one borrower into another. Everything about a crunch surviving a rate cut follows from that one sentence.

There is a second turn of the screw, and it is why raising the rate can make a careful lender worse off rather than better. A borrower who fully expects to pay every instalment cares a great deal about the rate. A borrower who privately doubts they will get through the year cares much less. So if the rate goes up, the applicants who accept it are disproportionately the ones who were never going to repay in full. So a higher rate quietly changes who is standing in the queue. The name for the effect is adverse selectionWhere the price of a deal changes who is willing to accept it, and changes it in the direction the other side does not want. Set out for credit markets by Stiglitz and Weiss in 1981., and the argument that a lender may therefore prefer to lend less rather than charge more was set out by Stiglitz and Weiss in 1981.

The household version is a landlord with a flat to let. A tenant who seems likely to trash the place does not become acceptable at a higher rent. Raising the rent gets rid of the careful tenants first and leaves exactly the applicants the landlord was worried about. At that point the landlord does not raise the rent again. The landlord leaves the flat empty and waits for a different applicant. Leaving it empty is refusing to trade at any price, and it is completely rational.

How can lending fall while the policy rate is being cut?

Here is the case from Sankhya. Taken as a pair, the two facts look at first like a contradiction. In the base case, Sankhya's credit stockThe total amount of borrowing outstanding across an economy at a point in time. The amount newly lent during a year is the flow instead. stood at Rs 14,00,000 crore and grew 10.00 per cent over the year, reaching Rs 15,40,000 crore. In the crunch case, the same starting stock grew 2.00 per cent, reaching Rs 14,28,000 crore. And across that same year the repo rateThe rate at which a central bank lends short term to banks against securities. The repo rate sets the level around which the market's overnight rates settle. was cut by 50 basis pointsA hundredth of a percentage point, used so that small rate moves can be stated without ambiguity. Fifty of them add up to half a point., from 6.00 per cent to 5.50 per cent.

Credit growth of 2.00 per cent and a policy rate coming down by half a point sat in the same year in Sankhya, and neither figure is a mistake. Finding them contradictory means quietly assuming that price is the only force operating in a credit market. Drop that assumption and the pair is ordinary: the price of credit fell, and the willingness to extend it fell by more.

Work the amounts rather than the rates. The amounts are where the size of it shows up. In the base case Sankhya's lenders extended Rs 1,40,000 crore of net new credit across the year. In the crunch case they extended Rs 28,000 crore. The difference, Rs 1,12,000 crore, is lending that did not happen, and it is 80.00 per cent of the base case. Credit was being made cheaper all the while, and one rupee went out for every five that went out the year before.

The last two rows of that table carry the argument. Read them across. Four of the money rows are identical between the two right hand columns. Every rupee figure a reader would normally quote is the same. The only rows that differ are the terms attached to lending and the answer to which side moved, and the second of those is not observed at all. Which side moved is inferred, and inferring it is the work.

How can a crunch be told apart from ordinary weakness?

Three tests, and the honest framing is that each is an observation rather than an opinion. The question is not what the analyst thinks happened. Each test points at something specific that is written down somewhere, and each of the three points at the supply side without proving it.

The first test asks whether the terms attached to lending moved as well as the amount. A price is only one of the things a lender can change. A lender can also ask for more collateralAn asset put up against a loan. The lender may claim and sell it where the borrowing goes unpaid. Demanding more of it tightens a loan without touching the rate., fund a smaller share of what the borrower wants, shorten the loan, or attach conditions that did not exist last year. A borrower who has stopped wanting to borrow does not cause any of those to change. So terms moving alongside the amount is a fingerprint of the lender's hand. The innocent explanation is that the mix of people applying got worse on its own, so the same lending standards produce tighter looking terms without anyone having changed their mind.

The second test asks where the fall sits. A nervous lender refuses the hardest applications first. So a fall landing hardest on the borrowers who are most difficult to assess, the ones without a long record, without much to pledge, in businesses whose prospects are hard to read from the outside, is the pattern a pullback in supply makes. If instead the fall is spread evenly across every kind of borrower, a general loss of appetite for borrowing fits better. The innocent explanation here is that hard to assess borrowers are often also the most fragile, so a weak year makes exactly those people postpone their plans first, entirely of their own accord.

The third test asks whether the fall outlasts a fall in the price of credit. If the price of credit has come down and lending has not recovered, something other than price is binding. The innocent explanation is the one already established: transmissionHow a policy rate move travels outward until it shows up in what borrowers and lenders actually face. The journey takes time, so an effect trails its cause. works with a lag, so lending failing to move immediately after a cut can simply be a cut that has not arrived yet rather than a cut that cannot work.

Not one of the three tests settles the question on its own, and each has an explanation that has nothing to do with lenders pulling back, so they are read together and they narrow the question rather than closing it. When all three point the same way, the case for a crunch is strong. When they point in different directions, the honest sentence is that lending fell and the cause is not established. Writing that down is perfectly respectable.

Why does lending stopping slow the wider measures of money?

The mechanism comes from the notes on how lending creates deposits. The bulk of the money in an economy was lent into existence rather than printed, and that is why the widest measure of money in Sankhya stands at five times the narrowest one. A loan does not move money from somebody who had it to somebody who did not. A loan creates a depositA balance held with a bank that the holder can spend. The wider measures of money count it, and the notes on the four measures set them out. that was not there before, on the other side of the same act.

Deposit creation runs backwards as readily as it runs forwards, so when lending slows, the creation of new deposits slows with it, and the wider measures of money grow more slowly than they otherwise would have. The Rs 1,12,000 crore of lending that did not happen in the crunch case is also Rs 1,12,000 crore of deposits that were never brought into being alongside it.

Now the limit. The argument establishes a direction and nothing else: how much the wider measures slow by, and for how long, does not follow from it. How far the effect runs depends on what else is happening to the amount of currency people are holding, what the other components of the wider measures are doing, and how long the pullback lasts. The notes on the four measures of money and the notes on how lending creates deposits are where those parts sit. Direction is a real finding. A magnitude asserted without the parts that produce it is not.

What can a rate cut not do?

A rate cut is a good instrument aimed at a specific thing. The cut lowers the price of credit, and through that it lowers what borrowers pay and raises what marginal projects can bear. The effect is real, and in most conditions it is enough. Working in most conditions is why the instrument exists and why it is reached for first.

A cut does not touch a lender's view of whether the money will be repaid. There is no channel from the policy rate to that judgement. The cut arrives at the price and stops, and if the thing holding lending down is a judgement about repayment rather than a judgement about price, the cut lands next to the problem rather than on it.

A lever that works with a lag also works under a condition, and the condition is that the price is what was binding, so an argument that treats a rate cut as sufficient has skipped the condition rather than tested it. The lag says when to expect an effect. The condition says whether to expect one at all, and the second question is the one that gets missed.

None of that is a criticism of the instrument, and it is worth saying so plainly. A cut that fails to lift lending in a crunch has not malfunctioned. The cut has been aimed at a price in a period when price was not what stopped, and that is a statement about the situation rather than about the tool. A spanner does not fail when the problem is a power cut.

What does an analyst actually watch when the lending figure drops?

An analyst who does this for a living starts somewhere much narrower than the word crunch, and usually never reaches the word at all. One series drops under either cause. The other mostly stirs when it is the supply side that has pulled away. So the first move is to set the two beside each other and refuse to read either one alone.

The first is the volume, the figure everybody already has. The second is the terms: how much collateral is being asked for against the same sort of loan, what share of a project a lender will fund, how long loans are being written for, how many conditions have appeared that were not there last year. Volume down with terms unchanged is the weak demand corner. Volume down with terms visibly tighter is the corner where the supply side is doing the work.

Volume alone puts a crunch and a slump in demand in exactly the same column, and it is the terms that separate them. An analyst who only tracks the amount lent has no way of telling one from the other.

A lender runs the same read from the other side of the desk, and a household can run a version of it too. If the bank that would have granted a loan last year now wants a larger deposit, a guarantor and a shorter tenure for the same amount, the price is not what changed. Something about how the lender is judging households of that kind has changed, and that is a fact about the lender's willingness rather than about the household's wish to borrow.

The Republic of Sankhya is invented.
Educational material. Not advice on any investment, tax, budget or market position.