Credit Growth: The Cleanest Read on Economic Momentum

Credit growth is the change in the stock of lending outstanding between two dates. Credit growing faster than the economy means something different from credit merely keeping pace, so the change is read against nominal output growth. The figure also arrives often and is revised little. Those two properties earn it the name clean, and the name is a claim about the data rather than about the economy.

Two facts about lending stand behind that answer. The first is that lending is what brings the wider money measures into being, so the lending stock and the money stock move together rather than independently. The second is that a fall in lending has more than one possible cause sitting underneath it, and the lending figure by itself cannot separate them. Credit growth takes that same lending, measures it as a rate of change, and sets it beside the output of the economy the lending goes into. The word clean survives that treatment in one sense and fails it in another, and the difference between the two senses is the whole of the subject.

What exactly is credit growth measuring?

The commonest error with this figure lives in the gap between the thing measured and the number quoted about it, so start with the thing being measured.

The Republic of Sankhya, an invented economy, holds a stock of lending outstanding. On any given day the clock can in principle be stopped, every rupee that has been lent and not yet repaid added up, and one number written down. At the first date in this lesson that number is Rs 14,00,000 crore. Twelve months later it is Rs 15,40,000 crore. Those two levels are the two things that exist.

Credit growth is not either of them. Credit growth is the change between them, expressed as a rate: Rs 1,40,000 crore more on a base of Rs 14,00,000 crore, or 10.00 per cent. Credit growth is the growth rate of a stock, so what gets published is the change while what exists is the level, and almost every mistake made with this figure comes from forgetting which of the two is being quoted.

The household version makes the shape obvious in about four seconds. A shopkeeper's borrowing from a lender stands at Rs 80,000/- at the start of the year and Rs 88,000/- at the end. The stock is Rs 88,000/-. The growth is 10.00 per cent. If somebody says the shopkeeper's borrowing grew more slowly this year than last, that is a statement about the second number and nothing whatever about the first, and the first is the one the shopkeeper has to service.

One more piece of care about what is inside the stock. The stock is a net figure, not a tally of new lending done. Fresh lending during the year pushes it up, repayments pull it down, and anything the lender stops counting pulls it down as well. So a year of heavy new lending alongside heavy repayment can produce a very small change in the stock, and the growth rate will report that small change honestly while saying nothing about the traffic underneath it. The gap between the traffic underneath and the net change is the door through which several things walk in and move the figure without moving the economy. The same netting is why the stock of lending sits close to, but is not the same thing as, the broad moneyThe widest of the standard money measures, counting currency held by people alongside the deposits sitting in banks. Which measure counts what, and why more than one of them exists at all, is worked through in the notes on the money supply. measure that lending brings into being.

Why is credit growth read against nominal output growth rather than on its own?

A growth rate quoted by itself is a number with nothing to lean on. Ten per cent is quick if the thing around it is standing still and sluggish if the thing around it is sprinting. So the figure gets a companion, and the companion is the growth of nominalMeasured in the rupees that actually changed hands during the period, with no adjustment for prices themselves having moved. The output measure, and the gap between a nominal and a real reading of it, both belong to the notes on output and national income, where each is built from scratch. output.

The reason it is that companion and not another is almost embarrassingly practical. The lending stock is counted in rupees of the period. Nominal output is counted in rupees of the same period. Neither has been adjusted for anything, so the two are the same kind of quantity and can be set on one scale without a translation step in between. Put a price-adjusted output figure beside an unadjusted credit figure and the comparison quietly breaks. One of them has had the effect of moving prices taken out of it and the other has not.

Work the Sankhya case. Credit runs from Rs 14,00,000 crore to Rs 15,40,000 crore, an addition of Rs 1,40,000 crore and a rate of 10.00 per cent. Nominal output runs from Rs 17,47,200 crore to Rs 18,17,088 crore, an addition of Rs 69,888 crore and a rate of 4.00 per cent. Credit added Rs 1,40,000 crore of rupees in the same period that output added Rs 69,888 crore. Lending grew by twice the rupees on a stock that started smaller, and that is what outrunning the economy looks like once it is put in rupees rather than in rates.

The comparison settled one thing and left another open. The comparison settled that lending expanded faster than the economy it was lending into. Whether a rate of credit expansion is appropriate is not a question a measure answers, so the comparison settled nothing about whether the gap is a good thing, a bad thing or a thing at all.

What does a rising credit to output ratio actually say?

One divided by the other gives the credit to output ratio, the form the comparison usually travels in. At the first date Sankhya's Rs 14,00,000 crore of lending sits against Rs 17,47,200 crore of output, or 80.13 per cent. At the second date Rs 15,40,000 crore sits against Rs 18,17,088 crore, or 84.75 per cent. The ratio rose by 4.62 percentage pointsPercentages carry their own unit for the distance between one and another. Sankhya's ratio climbing from 80.13 to 84.75 has travelled 4.62 percentage points, and saying instead that it climbed 4.62 per cent would name a much smaller and quite different thing..

Now the discipline, and it is the same discipline every ratio anywhere demands. Both parts moved, and a ratio never reveals which of its two parts did the moving. The lending stock went up. Output went up too. The ratio rose only because the first went up faster than the second, and the single number 84.75 carries no record of that at all.

Sit with how many different worlds produce a rising credit to output ratio. Lending accelerating while output holds steady produces it. Lending holding steady while output stalls produces it. Lending actually shrinking while output shrinks faster produces it too. Nobody expects that third case, and it is the one worth remembering. In that last world the credit to output ratio is climbing at the exact moment the stock of lending is getting smaller, and a reader who takes the rising ratio as evidence of expanding credit has been given the wrong answer by a number that was never lying.

A habit fixes this, not a formula. When a ratio moves, both parts have to be fetched before anything is said about it. The denominatorThe number underneath in a ratio, the one everything else is being divided by. Even when the number on top has not moved by a single rupee, changing the denominator changes the ratio. has a life of its own, and it is the half that gets forgotten, because the thing named in the headline is nearly always the half on top.

Why is credit growth called a clean read, and what is that claim about?

Now the title gets tested. Credit growth has a reputation among people who follow economies for being one of the cleaner things on the release calendar, and the reputation is deserved. The reputation is deserved for something narrower than most readers assume.

Three properties earn it. The figure is assembled from lending records rather than from a survey of people being asked what they did, so nobody has to be sampled or telephoned. The figure therefore arrives soon after the period it describes rather than after the long compilation that a broad measure of activity needs. And because it comes from records that were already being kept for other reasons, it tends to change very little once first published, where a first estimate of national output can be reworked more than once as better information arrives.

Every one of those three is a property of the data and not a property of the economy, so calling credit growth clean is a claim about how the number is produced and never a claim that the number answers the question brought to it. A measure can be prompt, stable and precisely defined and still be pointed at something other than what the reader wanted to know. A timely measure of the wrong thing is still the wrong thing, and it is more dangerous than a late one. It arrives first and sets the frame everything else gets read against.

A thermometer taped to the outside of a window makes the point. The thermometer reports quickly, it barely ever needs correcting, and it reports the temperature of the glass rather than of the room. Nothing is wrong with the instrument. Everything depends on knowing what it is touching. Credit growth touches the lending stock. Whether the lending stock is a good proxy for the momentum of an economy is a separate question, and it is the one worth asking seriously. Clean also carries no promise that the figure covers all the lending there is, and that turns out to matter a great deal.

What moves the credit growth figure when nothing has moved in the economy?

Whether the figure can be used or only quoted turns on what else is able to move it. There are at least four routine ways for credit growth to move by a visible amount while the economy it is supposed to be reading has done nothing at all.

The first is lending shifting between the measured and the unmeasured parts of a system. Every credit aggregate has a boundary drawn around it: certain lenders are inside the count and certain lenders are outside it. When borrowing migrates across that boundary in either direction, the measured stock changes and the total amount of borrowing in the economy does not. A borrower who moves from a lender outside the count to one inside it has changed nothing about how much is owed and everything about how much is counted.

The second is a single very large borrowing landing inside the period. One transaction of unusual size, arranged for reasons of its own, can lift the growth rate for a whole period by itself. Nothing broad happened. One deal happened, and the aggregate has no way of saying so.

The third is a write-off, and it is the one that catches people out because it runs the wrong way. When a lender concludes that an amount will not come back and stops carrying it, the stock of lending outstanding falls. No borrower repaid anything. No borrowing was declined. The economy is exactly where it was and the growth rate has been pushed down by an accounting decision about what to keep counting.

The fourth is a base effect, a property of the comparison rather than of the period. A growth rate is always measured against an earlier figure, so an earlier figure that was unusually high or unusually low bends this period's rate without this period having done anything unusual. If the base period happened to contain the second item on this list, the following period's growth rate will look weak purely because it is being measured against a bulge.

All four move the published figure while the economy stands still, so a reader who cannot name them will read noise as signal, and the four together are why the word clean had to be pinned to the data rather than allowed to float across to the economy.

What does a credit growth figure refuse to tell?

Even with all four of those distortions set aside, and a credit growth figure that is honest, timely and uncontaminated, there is still a whole category of question it does not answer, and being clear about the category is what separates using the figure from being used by it.

A credit growth figure is a volume, and a volume carries no information about quality: it does not say who borrowed, what the borrowing was for, or whether any of it will come back. Rs 1,40,000 crore was added to the Sankhya lending stock. The same addition is compatible with the money having gone to households buying homes, to businesses buying equipment that will produce something, or to borrowers covering short-term running costs after their receipts came in late. The three worlds behave completely differently in the years that follow, and the growth figure is identical in all three.

Draw the three worlds as bars and all three come out exactly the same length. Each is Rs 1,40,000 crore. Each is 10.00 per cent. The segments inside them are entirely different, and the published growth figure is the length of the bar rather than anything about the segments.

Volume and quality are two separate questions requiring two separate kinds of evidence, and a figure built to answer the first can never be squeezed into answering the second. Anyone who wants to know about quality has to go and get the composition of the lending, a different table, and then evidence about repayment behaviour, a different table again. Neither is inside the growth rate, and neither becomes inside it because the growth rate arrived first.

How does somebody with a decision to make actually use the figure?

Watch how an analyst covering an economy handles the figure. The handling teaches more than the definition does, and it looks nothing like reading a single number off a screen.

The first move is never to look at the growth rate alone. The growth rate goes straight beside nominal output growth: the pair says whether lending is outrunning the economy or keeping pace with it, and neither figure says that by itself. The second move is to look at the same figure against the same period a year earlier, a different comparison doing a different job. The year-earlier comparison catches a base effect by showing whether the period being compared against was itself unusual.

An analyst runs both comparisons because each catches something the other misses: the output comparison catches credit outrunning the economy and the year-earlier comparison catches a distorted base, and no single comparison catches both.

After that come the questions that matter. Which lending is inside the growth, meaning the composition. Did anything cross the measurement boundary during the period. Was there a write-off large enough to matter. Was there a single borrowing large enough to carry the rate on its own. Only then does the figure become usable, and by that point it has stopped being a headline and become the label on a folder that has to be opened.

A lender inside Sankhya arranges the same material differently again. Aggregate credit growth tells a lender roughly how crowded the market for borrowers is, and nothing about the specific borrower on the desk. A household feels a version of this too: what reaches a household is whether the loan it applied for was approved and on what terms, and the route from a national growth rate to that decision runs through a great many things that are not in the growth rate. None of the three treats the figure as a verdict. The figure is an input. Everything that would turn it into a conclusion lives in other tables and has to be gone and got. The policy rateThe rate the central bank sets, which anchors the short-dated rates across a system. Monetary policy and the rate corridor are covered separately. and the position in the business cycleThe repeating pattern of an economy speeding up and slowing down across several years. The economic cycle is built in its own notes, and the cycle is the very thing a credit growth figure is so often hoped to be reading. are two of those, and a shift of twenty five basis pointsA hundredth of one percentage point, which makes a hundred of them add up to a single per cent. Rate moves get quoted this way because saying a rate moved by a per cent leaves it ambiguous which per cent was meant. in a policy rate belongs to a different question from the one this figure answers.

What does the Sankhya case look like worked from end to end?

One worked case carries the whole argument, with every step laid out so it can be checked line by line. The same case is then produced a second time from different parts, and the second production proves the earlier claim rather than asserting it.

Every line reconciles. Rs 14,00,000 crore plus Rs 1,40,000 crore is Rs 15,40,000 crore, and Rs 1,40,000 crore on Rs 14,00,000 crore is 10.00 per cent. Rs 17,47,200 crore plus Rs 69,888 crore is Rs 18,17,088 crore, and that addition on that base is 4.00 per cent. Rs 14,00,000 crore divided by Rs 17,47,200 crore is 80.13 per cent, Rs 15,40,000 crore divided by Rs 18,17,088 crore is 84.75 per cent, and the gap between those two shares is 4.62 points.

Now produce the same 4.62 points from somewhere else. Hold the starting figures exactly where they are. Suppose credit had grown 7.88 per cent instead of 10.00 per cent, taking the stock to Rs 15,10,320 crore, and nominal output 2.00 per cent instead of 4.00 per cent, taking output to Rs 17,82,144 crore. Divide: Rs 15,10,320 crore over Rs 17,82,144 crore is 84.75 per cent. The ratio rose 4.62 points, exactly as before, out of a picture in which credit grew more slowly and the economy grew more slowly still.

Two different pictures produced one identical ratio move, so the 4.62 points on its own cannot distinguish an economy where credit ran at 10.00 per cent from one where it ran at 7.88 per cent, and that is the whole argument for never quoting a ratio move without both of its parts beside it.