India's Growth Model: What Actually Drives It

India's growth is usually explained by three structural features: services grew large before manufacturing had absorbed the workforce, demand comes mostly from inside the country, and informal arrangements remain widespread in how people work. Each feature has a mechanism behind it and a constraint attached to it, and together they explain why the same growth rate can produce very different employment. The magnitudes belong to the bodies that publish them.

A mechanism and a magnitude are two different kinds of claim. A mechanism can be reasoned through and survives the decade. A magnitude belongs to one particular year, to one particular definition, and to whoever compiled it, so a magnitude has to come from the body that compiles it. All the arithmetic below is carried by the Republic of Sankhya, an invented country standing in for a real one.

Underneath the answer sits one idea. A growth rate is a summary, and summarising means throwing information away. The subject here is what a growth rate throws away, namely what actually grewThe composition of growth: which parts of an economy account for the rise, rather than how big the rise was in total.. Two countries can post the identical figure and share almost nothing else. The figure does not supply the differences that survive it, so every one of those differences has to be reconstructed by hand.

Four ideas are already in place. The measure itself, and how much of a country's activity a compiled series manages to reach, is covered under the measurement of output. Output per worker, and why it is the one source of growth that can keep going indefinitely, is covered under productivity. Why work and output are two separate series rather than one is covered under the comparison of employment growth with economic growth. Work that moves out of unrecorded arrangements and into recorded ones is covered under formalisation. None of the four is defined again below. The four are picked up together and pointed at a single country.

Why does it matter what grew, when the growth rate is the same?

The mechanism is easier to see on a street than in an economy. Two neighbouring towns each add the same amount of new business in a year. In the first, a workshop opens on the edge of town that needs forty people standing at benches. In the second, a small design practice opens above the bank and needs six people with laptops. If either town kept accounts, both would record the same rise. Walking down the two streets in the evening, nobody would believe the same thing had happened. Forty households in the first town now have a wage that did not exist last year. Six do in the second.

An economy is that street repeated a few hundred thousand times, and a growth rate is what results when all of it is added up and printed as one number. A growth rate is a summary. Composition is what a growth rate summarises away, and composition decides how much work the growth produced. This is not a criticism of the measure. The measure was built to answer a different question, how much more was produced, and it answers that question well. A growth rate simply cannot answer the question a reader usually has next.

Every claim in this guide is a structural featureA lasting characteristic of how an economy is put together, as opposed to a reading for one year. A structural feature changes slowly and shapes what any year's figures will look like. rather than a reading for one year. The distinction between a structural feature and a one year reading is what makes the argument checkable. A share moves, a rank moves, a rate moves every quarter. The reason a workshop hires more people per rupee of output than a design practice does not move, and it can be reasoned about without looking anything up.

What changes when an economy reaches services before manufacturing has absorbed the workforce?

The route most large economies travelled runs farming, then factories, then services, roughly in that order and over a long stretch of time. The middle stage did something specific to the workforce, and it is worth being precise about what. A factory floor takes people whose formal schooling stopped early, gives them work that can be learned by doing it, pays a wage on a fixed day, and puts a name on a register. Each rupee of output that comes off that floor needed comparatively many pairs of hands to make it. So the middle stage is not just an economic stage. The factory stage is the mechanism by which very large numbers of people moved from unpaid or seasonal work into recorded, waged work.

Services led growthA pattern in which service activities account for most of the rise in output, rather than farming or factory production. describes a path that reaches the third stage while the second one is still thin. The output arrives. Service sectors differ enormously in how many people they need per unit of output, and the high value ones need very few. So the jobs arrive in a different quantity. A software practice, a design studio, a research desk and a specialist hospital department all produce a great deal of value per person, and every one of those people arrived already trained. The same growth therefore produces less formal employment on the services first path than the manufacturing first path would have produced, and that is a mechanism rather than a judgement about either path.

Three larger claims do not follow from this one. Services are not worth less than factory output, the output is not somehow unreal, and the path was not chosen badly. All three of those would be opinions rather than arithmetic. The claim is narrower and it is arithmetic: if a sector needs four times as much output per additional worker, then an identical rupee increment routed through it hires a quarter as many people. Nothing about that depends on which country or which year is in question.

The constraint this feature carries is that it concentrates the gains in the part of the workforce that already had skills. The design practice above the bank hires the person who finished a degree. The design practice does not hire the person who left school at fourteen and has been loading trucks since. Growth that runs through activities of that kind can be perfectly genuine and still leave a large share of the workforce exactly where it was, and that is the specific thing a reader should be alert to rather than a general worry.

What does it mean for growth to be led by demand from inside the country?

Two tailors, one street apart. The first sews for the households around him: school uniforms in June, wedding clothes in the season, alterations all year. The second sews shirts for a single buyer overseas who collects a container every month. In a bad year abroad, the first tailor barely notices and the second one closes his shutter. In a very good year abroad, the first tailor still barely notices and the second one doubles. Neither is the safer business in general. The two tailors are exposed to different things, and that is all.

An economy whose output is mostly bought by the people who live in it behaves like the first tailor. Domestic demandSpending by households, businesses and government inside the country, as against spending by buyers abroad. is the larger part of what keeps its production moving, so a slump in external demandSpending by buyers outside the country on what this country produces. External demand reaches the producer through exports. reaches it late, weakly, and through indirect channels rather than through a collapse in orders. The lateness is a real property, and it shows up in how such an economy behaves when the rest of the world has a bad year.

The same distance that stops a bad year abroad from arriving also stops a good one from arriving, so insulation is a two sided property rather than an advantage. This is the sentence readers skip. A country that cannot be knocked over by a downturn elsewhere also cannot be lifted out of a weak patch by a boom elsewhere. If domestic spending is soft, there is no external buyer standing by to take the surplus. Nearly all of the customers are inside the country, so the economy has to solve its weak year from the inside.

The constraint this feature carries is that domestic demand is bounded by domestic incomes. People can only buy what they can pay for. So an economy that grows mainly by selling to itself needs the incomes of the people inside it to keep rising, year after year, or the demand simply runs out of room. And incomes rise sustainably for one reason: each worker produces more than before. Domestic demand therefore drags the productivity question into the middle of the table, and productivity, rather than exports, is the subject that follows from it.

What does a largely informal workforce change first, the production or the measurement?

A woman runs a tailoring business from the back room of her house. Four machines, two people who come in daily, a steady stream of customers from the surrounding lanes. Real work. Real output. Real income for three households. There is no registration, no payroll, no filing, and nothing that will ever reach a statistical form unless somebody comes and asks. Multiplied by an enormous number, she is what people mean by informal workWork carried on outside registered arrangements: no formal contract of employment, no payroll record, and usually no coverage by workplace protections..

Informality does not mean the work is unproductive, or illegitimate, or invisible to the people doing it. The clothes exist. The money changes hands. Informality changes what the statistics can see first, long before it changes anything about what gets made. Production and measurement are two separate things, and it is the second one that informality reaches first.

Four consequences follow. They get run together, so they are worth separating one at a time. Work outside recorded arrangements is harder to measure, so an output series built from registers and returns will reach only part of it. The transaction leaves no trace a tax system can find, so such work is harder to tax. A lender wants a documented income and there is none, so such work is harder to lend against. Workplace rules attach to a recorded relationship between an employer and an employee, and no such relationship exists here, so such work is harder to protect.

The constraint this feature carries is that measured growth and actual growth can pull apart during any period in which the boundary between the two is moving. If activity that already existed crosses into the recorded part, the published series rises without one extra item being produced. The boundary effect is not a footnote. Formalisation works it all the way through with the arithmetic, and settles how large the distortion can get.

What happens when a services first path meets an informal workforce?

The third feature does something to the first one that neither does alone. Put together, they produce a specific and very deceiving pattern: measured output per worker that looks strong, for the plain reason that the part of the economy being measured is the more productive part of it.

Take Sankhya, the invented country this sequence uses, and add one assumption to it used only in this block. Suppose Sankhya truly produces Rs 20,00,000 crore of output, that the compiled series reaches Rs 16,00,000 crore of it, and that the remaining Rs 4,00,000 crore sits outside the record. Suppose 8.00 crore people work, that the work of 4.00 crore of them shows up in the record, and that the work of the other 4.00 crore does not. Now divide, and watch what happens.

One economy, three answers, and every one of the three is arithmetically correct. The first row says Rs 4,00,000 and describes the recorded part. The second says Rs 2,50,000 and describes the country. The third says Rs 2,00,000 and describes nothing at all. The third row puts recorded output over everybody who works, so the top of the fraction and the bottom cover different things. The first figure is 1.6 times the second and twice the third, and none of that spread came from anybody producing anything differently.

The spread is a composition effectA change in an average that comes from a shift in which items are included in it, rather than from any change in the items themselves., and reading that spread as a productivity gain across the whole economy is a category error rather than a small overstatement. The average did not rise because anyone got better at anything. The average rose because the pool being averaged is not the pool the reader had in mind. A school whose average marks rise after the weakest class stops sitting the exam has not taught anybody better, and the arithmetic here is the same arithmetic.

Why does the services first path make this worse rather than better? Because the recorded, high output activities and the unrecorded, low output ones sit at opposite ends of the same scale. The wider that gap, the more the recorded average overstates the whole. An economy where the recorded and unrecorded parts produced similar amounts per person would barely show the effect at all.

One question answered three times, on an invented economy.

The argument here is about which parts moved, not about how much a total moved, so the useful exercise is to hold the total still and let the parts vary. The arithmetic below belongs entirely to the Republic of Sankhya, a stand in chosen so that the reasoning stays visible.

The question is one a reader genuinely has. A year of growth arrives. How much recorded, formal work does it produce? Sankhya starts the year with a recorded output of Rs 16,00,000 crore and grows 5.00 per cent, adding Rs 80,000 crore. The extra output is held fixed through all three answers below. Two assumptions carry the employment arithmetic, and both are the panel's own inventions rather than facts about anywhere: every Rs 1,00,000 of extra manufacturing output needs one more worker, and every Rs 4,00,000 of extra services output needs one more worker.

Path A worked by hand once makes the rest checkable. Three quarters of Rs 80,000 crore is Rs 60,000 crore. At one worker for every Rs 1,00,000 of output, that is 60,00,000 people, or 0.60 crore. The remaining Rs 20,000 crore of services output, at one worker for every Rs 4,00,000, is 5,00,000 people, or 0.05 crore. Together that is 0.65 crore of new work, and three fifths of it recorded is 0.39 crore. Path B moves the same rupees the other way and lands at 0.35 crore of new work and 0.21 crore recorded. Path C keeps B's composition and changes only how much of that work sits inside recorded arrangements, landing at 0.07 crore.

The growth rate is 5.00 per cent in all three, the extra output is Rs 80,000 crore in all three, and the recorded new work runs from 0.39 crore down to 0.07 crore, a spread of more than five times. Nobody grew faster. Nobody produced more. The only thing that changed between the first answer and the third is which activities the rupees passed through and how much of the resulting work sits inside recorded arrangements. The whole argument sits in that one table.

The actual magnitudes for India, for every quantity the walk has given a Sankhya version, exist and are published by named bodies. The section below says which body publishes which kind of thing.

What goes wrong when two countries report the same growth rate?

Here is the mistake in its natural habitat. Somebody puts two countries side by side, reads the growth rate for the same year off each, sees that the two match, and concludes that the two economies are performing similarly. Every step of that is arithmetically fine. The figures are correctly compiled and correctly compared. The conclusion is still wrong, and it is wrong because a growth rate was never the kind of thing that could support it.

What does an analyst actually do with a structural reading?

An analyst covering companies in a country does not use a structural reading to forecast the country. The analyst uses a structural reading to work out which of a company's numbers are exposed to which conditions, a much smaller and much more useful question.

Take two businesses in the same city, on the same exchange, in the same reporting currency. One sells packaged food into shops across the country. The other writes software for clients in three foreign markets. On a structural reading they are not in the same economy at all for the purpose of forecasting either one. The first business rises and falls with domestic incomes, so the analyst watches wage growth, employment and how much of that employment is in recorded, waged arrangements. Recorded income is the income that gets spent through the formal shops the company sells to. The second business rises and falls with conditions in three other places, so the domestic picture tells the analyst comparatively little about its next four quarters.

A business selling into domestic demand and a business selling abroad are not in the same economy for forecasting purposes, even when they are headquartered on the same street. That is the practical payoff of a structural reading and it requires no figure at all. A credit officer uses the same reading differently: a borrower whose customers are all local is exposed to a domestic slowdown and insulated from a foreign one, and the covenants and the stress case should reflect which of the two the lender is actually underwriting.

The one more thing a careful reader does with the informality feature is to treat a documented income as a smaller category than an actual income. A household running a business from a back room has an income and no documents. A lender who works only from documents is not looking at a poorer market. The lender is looking at a smaller slice of the same market, and knowing about the slice is a different thing from measuring it.

Where are the actual numbers for India found?

The mechanism ends here, and a magnitude is usually what a reader wants next. A magnitude arrives attached to three things that never travel with it when it is copied elsewhere: the definition it was built on, the coverage it claims, and the revision history that says how much it has already moved since it was first published. So the right way to get one is to go to whoever compiles it.

Three bodies are worth naming for the questions raised here. The Ministry of Statistics and Programme Implementation is the issuer for India's national accounts, the series in which output for the country is compiled and published. The National Statistical Office carries out the survey work behind India's employment statistics, the material standing behind any question about how many people work and under what kind of arrangement. The Ministry of Finance issues the Economic Survey, a document that gathers the statistical picture together and discusses the same structural questions. Each of the three is a primary sourceThe body that compiles and issues a figure itself, as opposed to anyone who quotes, summarises or reprints it afterwards. for what it puts out, and a primary source is the only place a magnitude comes with its definition and its date attached.

An explanation that carried Indian magnitudes would stop being an explanation and start being a record somebody has to maintain, and a record nobody maintains goes wrong quietly rather than loudly. A mechanism does not go stale. A figure printed in the wrong decade sits there looking exactly as confident as it did the day it was correct, and the confidence is what makes it dangerous rather than merely unhelpful.

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