The Formalisation of the Indian Economy

Formalisation is economic activity moving inside the reach of the systems that record it: a business registered, a wage written down, a payment that leaves a trail. Formalisation changes what the statistics can see before it changes what gets made. A formalising economy can therefore print growth it did not produce, and reading such a figure correctly is a skill in its own right.

The answer rests on two things worth naming before anything is built on top of them. The first is output itself, and how a recorded output series is put together, including the plain fact that such a series cannot count what never reaches it. The second is the part of India's workforce that operates outside registration, employment records and filed accounts, a lasting structural feature rather than a temporary one. Both are settled. One job remains: working out what happens to a growth rate when the reach of the recording system itself is changing. FormalisationActivity moving inside the reach of the systems that record economic life, chiefly business registration, employment records, banked payments and filed accounts. is the name for that change in reach.

The arithmetic below runs on the Republic of Sankhya, an invented mixed economy of some 20 crore people. Indian magnitudes sit with the issuers who maintain and revise them. A quantity lifted away from its issuer goes wrong on the day it is revised, so the magnitude belongs with whoever publishes it and the mechanism is what travels. Five things are at stake: what formalisation is without using the word, how a published rise splits into the part produced and the part merely counted, how far a growth rate can be pushed by the boundary alone, what a business trades away by crossing, and the one question any growth figure from an economy with a rising recorded share has to answer.

What does formalisation actually mean?

Formalisation is a line moving, not a state of being. Draw a line through an economy. On one side sits everything the recording systems can see: businesses on a register, wages on a payroll, sales through a payment rail that keeps a record, accounts that get filed somewhere. On the other side sits work that is just as real, producing goods people actually eat and services people actually use, but leaving no trace those systems can pick up. The line just drawn is the measurement boundaryThe edge of what a statistical system can observe. Activity inside it can be counted; activity outside it exists but cannot enter the published totals., and formalisation is the movement of that line across activity that was already there.

Nothing about the underlying activity has to change for something to cross that line, and that single observation carries everything that follows. Think of a tailor who has stitched from a room at the back of her house for eleven years. Same two machines, same customers, same volume of work, same hours. In the twelfth year she registers the business and starts taking payment into a bank account. Not one extra shirt was made. Not one extra rupee of value was created that week. But an economy that could not see her at all can now see her completely, and every total that describes that economy has just got larger.

What sits on each side of the line?

On the recorded side, four markers do almost all the work: a registration somebody can look up, employment that exists as a record rather than as an understanding between two people, payments that move through something which keeps its own memory of them, and accounts that get filed and can therefore be read by a stranger. Any one of those makes a business partly visible. All four together make it fully countable, and full countability is what the recorded economyThe portion of economic activity that reaches the statistical system through registrations, employment records, banked payments and filed accounts, and so can enter published totals. means in practice.

On the other side sits work with none of those markers, and here is where a reader most easily goes wrong. Informal does not mean unproductive, unskilled or illegitimate, and treating it that way misreads the majority of economic life in a great many countries. Informal workWork that produces real output but leaves no registration, payroll record, banked payment or filed account behind it, so no statistical system can observe it directly. feeds people, moves goods, builds houses and repairs everything. Picture the caterer who hires forty people for three days of a wedding, feeds six hundred guests, is paid in cash by the bride's household, and pays each of the forty by hand on the last night. Six hundred meals were genuinely produced. Forty people genuinely worked. Not one line of that reaches any record, and no counter anywhere in the country can see it happened.

Why does measured growth rise when nothing new is produced?

Because the published series counts the recorded side, and only the recorded side. The sequence is worth following slowly. A statistical system builds its total from what it can observe. Something crossing the boundary was not in last year's total and is in this year's total. So the total rises. Comparing the two totals gives a growth rate that is partly real production and partly the boundary sweeping up activity that was there all along. Measured growthThe change from one published output total to the next. The rate moves with real production and also with any change in how much of the economy the recording system can reach. is therefore the sum of two quite different things, and only one of them is production.

The instinct is to call this a fault, and the instinct is wrong. Nobody has miscounted anything. Both totals correctly report what was inside the boundary at the time, so the gap is not a data error. Nothing was misstated by anybody, so the gap is not fraud. An office can only ever count what reaches it, and an office that guessed at what did not reach it would be doing something far worse, so the gap is no fault of any statistical office. The gap is what happens when the thing being measured and the ability to measure it change during the same period. Any measure has this property. A person weighed on Monday in a shirt and on Friday without one is weighed honestly both times. The difference says nothing about the body.

The practical consequence is uncomfortable and worth stating flatly. For as long as the boundary is moving in one direction, measured growth runs above true growth, every year, without a single year of deliberate misreporting. The boundary is not wandering randomly, so the gap is not noise that averages out. The boundary drifts one way, and the error drifts one way with it.

How large is that gap, worked on one invented economy?

Claims about a mechanism are cheap until somebody puts numbers on them, so here is the whole thing in one arithmetic, on the Republic of Sankhya.

Sankhya produces Rs 20,00,000 crore of true output in year one. Of that, Rs 16,00,000 crore is recorded, or 80.00 per cent, and Rs 4,00,000 crore is not. The Rs 16,00,000 crore is the only part anybody can see, so it is the only part published. Now let year two be an ordinary year in which true output rises 3.00 per cent, to Rs 20,60,000 crore. Hold the boundary still for a moment. The recorded part would be Rs 16,48,000 crore and the unrecorded part Rs 4,12,000 crore, and the published series would rise by exactly 3.00 per cent, matching the truth.

Now let the boundary move. A quarter of that unrecorded part, Rs 1,03,000 crore, crosses into the recorded economy during year two. Recorded output is therefore Rs 16,48,000 crore plus Rs 1,03,000 crore, or Rs 17,51,000 crore, and the unrecorded part falls to Rs 3,09,000 crore. Nothing was created and nothing was destroyed, so the two still sum to Rs 20,60,000 crore. Measured growth is Rs 1,51,000 crore on Rs 16,00,000 crore, or 9.44 per cent, against true growth of 3.00 per cent: a gap of 6.44 percentage points, and not one rupee of extra output was produced to create it.

Split the published rise and the point lands hard. Of the Rs 1,51,000 crore that the recorded series went up by, Rs 48,000 crore is genuine growth on the part that was already being counted, and Rs 1,03,000 crore is output that existed in year one and merely changed sides. The crossing accounts for 68.2 per cent of the entire published increase. More than two thirds of what a reader would call this year's growth was produced before the year began.

Now run year three, and change one thing only: leave the boundary exactly where year two left it, at 85.00 per cent of true output. True output rises another 3.00 per cent, to Rs 21,21,800 crore. Recorded output rises to Rs 18,03,530 crore, and unrecorded to Rs 3,18,270 crore. The published growth rate is 3.00 per cent, and the true growth rate is 3.00 per cent. The two rates coincide the moment the boundary stops moving. The gap was never a property of the economy at all, only of the change in what could be seen.

With the selector at a fall of 2.00 per cent and the crossing left at a quarter, the case worth remembering appears. True output falls to Rs 19,60,000 crore. The recorded part before any crossing is Rs 15,68,000 crore, the unrecorded part is Rs 3,92,000 crore, and a quarter of that, Rs 98,000 crore, crosses. Recorded output is Rs 16,66,000 crore. The published series prints growth of 4.13 per cent for a year in which Sankhya actually shrank by 2.00 per cent, and every number in that series is correctly counted. The movement required is smaller than instinct suggests: with true output down 2.00 per cent, a crossing of just 9 per cent of the unrecorded part is enough to turn the printed rate positive.

What actually changes for a business that crosses the line?

Everything about how the business is seen, and rather less about how it works. Start with what it gains. The business becomes visible to lenders, the largest single change of all, and lending is taken up below. Buyers who cannot buy without an invoice, meaning most large buyers and every government one, come into reach, so a ceiling on customer size lifts. Records now outlive the person holding them in his head, so the business becomes something that can be handed over, sold or borrowed against. And a structure now does some of the trusting, so hiring can go beyond the circle of people the owner already trusts personally.

Then come the losses, and they are just as real. Being visible to the recording systems means being visible to the ones that levy, so costs rise. Obligations to workers become enforceable rather than customary, and enforceable obligations cost money. Record keeping consumes time and often fees that the business did not previously spend at all, and for a small business that burden is not trivially small. Everything documented is a commitment somebody can hold the business to later, so flexibility narrows.

The trade runs both ways, and that is exactly why crossing the boundary is a decision the business makes rather than an inevitability it undergoes. A reader who treats formalisation as obviously and immediately good for every business that does it has not looked at the second column. A reader who treats it as pure cost has not looked at the first. A particular business tips one way or the other on how badly it needs something in the first column that it cannot be had any other way, and that something is almost always credit accessWhether a business can borrow at all, and at what price. Access depends less on how good the business is than on how much of it a lender can verify from the outside..

What does formalisation change for a lender?

Formalisation changes the category the borrower sits in, and the change of category is bigger than it sounds. A lender facing an unrecorded business is not looking at a bad borrower. The lender is looking at a business it cannot assess at all, and a bad borrower and an unassessable one are two entirely different problems. A bad borrower has records, and the records say something a lender does not like: thin margins, a stretched position, a history of late payment. Thin margins are information. The lender can price them, shorten the tenor, ask for security, or decline on evidence. An unrecorded business hands over nothing to be liked or disliked. The file is not bad. The file is empty.

The lender's response to an empty file looks like harshness and is mostly arithmetic. A lender prices for the risk it can measure and for the uncertainty it cannot. When there is nothing to measure, all of it is uncertainty, so the price has to carry the whole of it, and quite often the price that would carry it is one no borrower would accept, at which point the answer is simply no. An unrecorded business is declined or charged for the not-knowing rather than judged to be a poor risk, and formalisation is what moves it out of that second category into the first. That change in assessabilityWhether an outsider can form a view of a business at all from evidence the business can hand over. Assessability is separate from whether the view, once formed, is favourable. is the single largest consequence of the boundary moving for anybody actually lending money.

The same thing is visible at household scale. Two men want to borrow Rs 3,00,000 to buy a delivery van. The first shows two years of bank statements and an income record: the lender reads them, dislikes some of what it sees, and offers a smaller amount at a higher rate. The second earns more than the first, every rupee in cash, and can show nothing at all. He is the better business and the worse application, and explaining that changes nothing the lender can verify. The outcome says nothing about either man. The outcome says what a lender can see, and that limitation is the statistical system's own, arriving at a different door.

What should be done with a growth figure from a formalising economy?

One question comes before anything else: how much of this rise is the boundary moving? The question does not have a clean answer, and pretending otherwise would be the second mistake stacked on the first. Estimating how much activity crossed in a given year means estimating the size of something that by definition was not observed. Serious estimates therefore differ from each other, sometimes by a lot, and they get revised. There is no arrangement of arithmetic that turns an unobserved quantity into a precise one.

So the honest procedure is not a formula but a posture. If a credible estimate of the boundary movement exists, take it off and then say out loud that what remains is an estimate carrying a range, not a figure carrying a decimal. If no such estimate exists, hold the published rate more loosely and go looking for series the boundary cannot move: physical volumes, freight tonnage, power consumption, employment counts. Those series have problems of their own, but different ones, and agreement between two measures that fail differently is worth more than precision in one. The reader who says the growth rate is uncertain by an amount they cannot pin down has understood the situation better than the reader who quotes it to two decimal places. That is a statement about coverageHow much of what actually happened a statistical series manages to reach. Two series can cover different amounts, so both can be accurate about what they cover and still disagree., not about anybody's competence.

Where are the actual numbers for India published?

Every quantity worth having sits with an issuer who maintains it, revises it, and publishes what it covers alongside it. A figure lifted out of that setting loses the revision, loses the coverage note, and becomes wrong on a date the reader would never know about. The issuers are named below.

One more thing worth saying plainly about that routing. A published figure and the coverage note attached to it deserve the same attention. The difference between the two is the whole of formalisation. A number and a statement of what the number reaches are not separate pieces of information. The second is what settles whether the first can be compared with the one from three years ago.

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