Infrastructure-Led Growth in India: What the Claim Needs

Infrastructure-led growth is the claim that spending which leaves lasting assets behind raises what an economy can produce later, and not only what it demands this year. Two claims sit stacked inside that one phrase. The first is true of any spending at all. The second is the one worth arguing about, and it needs four separate things to be true before it holds.

Four ideas do the work below, and every one of them is built elsewhere. The split between spending that leaves an asset and spending that does not is settled under the classification of government spending, along with the warning that came attached to it. The three deficits are settled under the deficit measures. A stock of public debtThe accumulated amount a government still owes from past borrowing, as against the amount it borrows in any one year. Measured and discussed at length under public debt., and the way that stock is measured, is settled under public debt. So is the direction that heavy government borrowing pushes the price of funds.

The phrase itself turns up constantly in commentary, in budget coverage and in research notes, and it carries a working part and a decorative part that are easy to mistake for one another. Once the two are pulled apart, what would have to be true for the working part to hold can be listed, and the places the argument comes apart can be counted.

A reader meets this claim in India, where infrastructure investment is a standing subject of argument. The Union Budget is a real published document with a capital line in it. The Ministry of Finance prepares it. The Comptroller and Auditor General examines what was actually spent against what was authorised, and the Reserve Bank of India publishes compilations covering government finances. Each of those is a real body or a real document. The outlays, the projects, the targets, the periods and any assessment of whether a programme has worked belong to those published sources.

Every rupee amount here belongs to the Republic of Sankhya, an invented country. Its government account is the same one used throughout: receipts of Rs 3,00,000 crore set beside spending of Rs 4,00,000 crore, with Rs 70,000 crore of that spending sitting on the capital line, leaving a fiscal deficit of Rs 1,00,000 crore, against a stock of debt of Rs 10,00,000 crore carrying an average cost of 9.00 per cent. Sankhya output is Rs 17,47,200 crore. Every one of those figures divides cleanly, and no real fiscal account ever does.

What is actually being claimed when somebody says infrastructure-led growth?

Two things at once, and they are so different in strength that hearing them as one sentence is the first mistake. The two claims have to be pulled apart before anything else can be done with them.

The first claim is about demand. The government spends Rs 70,000 crore, and that money goes to contractors, to steel suppliers, to cement plants, to the people who operate the machinery and to the shops those people buy from on a Saturday. Orders are placed, wages are paid, and activity this year is higher than it would otherwise have been. The demand claim is true, and unremarkable. The demand claim is true of every rupee of government spending, including the rupees that leave nothing at all behind.

Test that against Sankhya. Revenue spending is Rs 3,30,000 crore, and Rs 90,000 crore of it is interest. Interest goes to whoever holds the government's earlier borrowing, and they in turn spend or lend it. Interest is demand this year, exactly as the capital line is demand this year. Nobody calls interest payments a growth strategy, and the reason is not that they fail the demand test. Interest payments pass the demand test. Interest payments fail the second test.

The second claim is about supply. The supply claim says that when the money has been spent and the dust has settled, something is standing there that was not standing there before, and that the something makes the country able to produce more than it could before. A bridge where there was a ferry queue. A transmission line where there was load shedding. A port berth where ships used to wait offshore. The supply claim is about what an economy is capable of, not about what it happened to want in one year, and it belongs with potential outputWhat an economy is able to turn out on a sustained basis given its workers, their skills, its machines and its infrastructure, as against the amount it happened to turn out in one particular year. Built earlier in this reading order and only borrowed here. rather than with the year's activity.

Here is the everyday version. A household spends Rs 40,000 on a wedding feast and a household spends Rs 40,000 on a borewell. Both are demand: the caterer earns, the driller earns, the money circulates identically. But a year later, the feast is a memory and the borewell is still pumping water, and the second household can now grow a crop it could not grow before. Both arrive in the accounts as spending of Rs 40,000, and so a confusion nobody would fall into at home happens constantly in fiscal commentary.

Why is the capital line the line this claim is made about?

Because it is the only line in the account defined by what survives the spending. The split between revenue and capital spending is set out under the classification of government spending. The consequence is what counts: talking about assets raising future capacity means talking about the spending that produces assets, and in a government account that is the capital line.

In Sankhya, capital spendingThe part of a government's spending that leaves something lasting behind, as against the part that pays for the running of things. Defined and worked through earlier in this reading order and only borrowed here. is Rs 70,000 crore where the whole of spending is Rs 4,00,000 crore. The share works out directly: 70,000 divided by 4,00,000 is 0.175, so 17.50 per cent. Against output of Rs 17,47,200 crore the same Rs 70,000 crore is 4.01 per cent. Both numbers are worth holding. The first shows how much of the government's own spending could even be a candidate for the claim, and the second shows how big that candidate is next to the economy it is supposed to move.

The warning arrives with the definition itself. The classification records what remains after the money was spent, and it says absolutely nothing about whether the thing was worth building. A bridge to a village of forty households and a bridge on a corridor carrying a quarter of the region's freight are both capital spending. A road built to a standard that survives two monsoons and a road built to a standard that survives one are both capital spending. The line in the account cannot see the difference, and it was never designed to.

So being on the capital line is necessary to the claim and nowhere near sufficient. Spending that leaves nothing behind cannot raise capacity by leaving something behind, so the capital line is necessary. The classification is a description of form and the claim is an assertion about consequence, and no amount of the first ever adds up to the second, so the capital line is nowhere near sufficient.

What has to be true before the supply claim actually holds?

Four things, in order, and each of them is a separate event that can fail on its own. Setting them out as conditions rather than as an argument is deliberate: conditions can be checked one at a time, years apart, by different people, without anyone having to forecast anything.

The first condition is that the money is spent rather than merely authorised. An amount appearing in a budget document is an appropriationPermission granted by a legislature for a stated amount to be spent on a stated purpose. Permission to spend and money actually leaving the account are two different events, and the second is what shows up in an audited outturn., which is permission, not payment. Permission and payment part company all the time: approvals arrive late in the year, land is not in hand, clearances are pending, and the year closes with a chunk undrawn. The reason the Comptroller and Auditor General exists as an institution is precisely that somebody has to check the outturn against the authorisation, and that is the check the first condition needs.

The second condition is that the asset is completed rather than started. Half a bridge carries nobody. Half a transmission line moves no power. A partly built asset has consumed the money, generated the demand, and delivered exactly none of the supply effect, and the reason this failure is so common is that it is invisible in an annual account: the spending appears, on time, in full, and the capacity does not.

The third condition is that the completed asset is used. A finished road with no traffic on it is a monument. The third condition sounds absurd until the list of things that have to line up for use to happen is set out: the connecting stretches at either end, the tariff or toll that makes using it worthwhile, the industries that would have generated the freight, the last stretch of wiring from the substation to the households. Capacity that exists and sits idle is capacity the economy is not producing with.

The fourth condition is that the use relieves something that was actually holding production back. This is the subtlest one and the easiest to skip. If lorries were queueing eleven hours at a checkpost and the new corridor cuts that to two, something real was in the way and is now less in the way. If lorries were moving freely and a parallel corridor is built beside the old one, nothing was binding, so the asset is completed and used and has relieved nothing. The everyday version is a household with one bathroom and six people building a second bathroom, against the same household building a third when the second already had no queue. Same spending, same completed asset, entirely different consequence.

Four conditions in sequence means four separate places the chain can break, and a claim resting on a chain is always weaker than each of its links sounds. Sit with that for a moment. The chain is the mathematical shape of the whole argument. Each link can be individually plausible and the chain can still be unreliable, and anyone who has waited on four separate approvals for one thing already understands this in their bones.

Who is paying for the asset while the capacity is still being built?

Somebody is, and the claim almost never mentions it. The arithmetic is the same in every account, so trace it through Sankhya.

Sankhya takes in Rs 3,00,000 crore and spends Rs 4,00,000 crore. The gap between those two is the fiscal deficitThe shortfall between a government's receipts and its spending across a single year, closed by borrowing. Built and set against the other deficits earlier in this reading order., Rs 1,00,000 crore, and that whole gap is borrowed. The capital line is Rs 70,000 crore. Since Rs 70,000 crore is less than Rs 1,00,000 crore, every rupee of the capital spending sits inside the borrowing, and there is Rs 30,000 crore of borrowing left over on top of it.

The leftover Rs 30,000 crore is exactly the revenue deficit, and the reason repays a moment of precision. Take Rs 70,000 crore out of Rs 1,00,000 crore and Rs 30,000 crore remains, and Rs 30,000 crore is the revenue deficit figure. The match is an identity and not a check. An identity holds by construction wherever no non-debt capital receipts exist, so the second route can never disagree with the first, and calling the match a confirmation would be false. The identity shows something narrower and genuinely useful. Borrowing exceeds capital spending by exactly the revenue deficit, so Rs 30,000 crore of what was borrowed funded consumption and left no asset behind at all.

Now attach the cost. Sankhya debt of Rs 10,00,000 crore at an average cost of 9.00 per cent produces the interest bill of Rs 90,000 crore. Heavy borrowing also presses on the price of funds in the market, and the direction of that press is set out under government borrowing and the price of funds. Carry the Rs 70,000 crore capital line at 9.00 per cent and it costs Rs 6,300 crore every year, for as long as it is outstanding, starting immediately.

So the claim, stated honestly, is that what the assets add to what the economy can produce outruns what the borrowing costs to carry, and no general argument settles that comparison for any real case. It cannot. Settling it needs a measured addition to capacity, a measured cost, and some account of how the economy would have fared had the spending never taken place, and not one of those three is available from a budget document. The comparison is an empiricalSettled by evidence collected from the world rather than by reasoning from definitions. An empirical question can be argued about correctly for years and still not be closed. question, and the honest position is to state it and leave it open rather than to resolve it in a confident voice.

Why can no single period settle the argument?

Because the two effects arrive on completely different clocks, and a year is long enough to contain the whole of one and only a sliver of the other.

The demand effect is fast and loud. The money is spent, the orders are placed, the wages are paid, and it shows up in the activity of that period in a way that is easy to measure and hard to miss. If Rs 70,000 crore goes out this year, something in this year's numbers moves, and everyone can see it move.

The supply effect is slow and quiet. The asset takes years to finish. Once finished, it takes more years for freight to reroute onto it, for the industrial units to be set up near it, for the households to connect to it. And all through those years, everything else in the economy is also moving: the weather, the price of imported fuel, the cycle, the policy rate, world demand, a hundred things at once. By the time the capacity effect is large enough to see, it is thoroughly mixed with everything else.

The mixing is the real obstacle, and it has a name worth knowing. Attributing a change to the asset would require a counterfactualA statement of what would have happened if the thing in question had not been done. A counterfactual cannot be observed, only estimated, and that is why attributing any outcome to any single cause is genuinely hard., meaning an account of what output would have been had the corridor never been built. Nobody can observe a counterfactual. A counterfactual can only be estimated, and reasonable people estimate it differently, so careful researchers spend years on a single corridor and still report ranges rather than answers.

The two effects run on different clocks, so a single period cannot separate them, and that is the structural reason infrastructure-led growth is argued about rather than settled. The people arguing are not careless. The evidence they need arrives slowly, arrives mixed, and never arrives in the tidy annual form that budget commentary runs on.

Can any general argument say whether it has worked in India?

No.

India runs infrastructure investment as a stated approach, and the numbers behind it sit in the Union Budget documents prepared by the Ministry of Finance, in the outturn and audit reports of the Comptroller and Auditor General, and in the compilations published by the Reserve Bank of India. Each of those bodies and documents is real, and the evidence lives in what they publish. The outlays, the projects, the targets, the periods and any verdict on whether a programme has worked all belong to those sources. An answer to that last one needs three things at once: an outlay, a measured result, and some defensible view of how matters would have gone in the absence of the spending.

Pointing at the sources rather than quoting them has two reasons behind it. The first is practical. Anything live keeps moving while written material stands still, so the moment an allocation is revised, whatever had quoted it has quietly begun to mislead, and whoever trusted it ends up worse informed than whoever was simply pointed at the source. The second reason is the more important one. The conditions that would have to be true are teachable, and whether they were true in a particular case is a matter of evidence, to be read in the sources themselves.

Consider what the four conditions provide. Any announcement about infrastructure and infrastructure-led growth, in India or anywhere else, can be split at once into its demand half and its supply half. The four conditions then ask whether the money was spent or only authorised, whether the assets were completed or only started, whether they are used or standing idle, and whether the use relieved something that was genuinely in the way. One more question asks what the borrowing costs to carry, and somebody who has quietly skipped that part becomes easy to notice. The questions are permanent, and none of them expires when a figure is revised.

What does somebody reading a real account actually do with this?

A claim like this almost never fails everywhere at once, so the reader stops treating it as one thing and starts hunting for the link that is broken. The claim fails at one point, and knowing which point is the whole of the useful information.

A credit analyst looking at a state or a public entity that has borrowed to build asks the first condition hardest. Was the money drawn down, and how much of the authorised amount closed the year unspent? The gap between authorisation and drawdown is visible in an outturn, it is exactly what an audit report examines, and it shows whether the borrowing is buying anything yet.

An equity analyst covering construction, cement or capital goods lives on the second and third conditions. Completion is what turns an order book into a settled receivable, and use is what determines whether the next round of orders follows the first. A pipeline of started and unfinished work is a very different thing from a pipeline of completed and running work, even when both are described with the same word.

The fourth condition is the one nobody checks, so somebody working on policy or economics pushes hardest there. Was anything actually in the way, and is it less in the way now? The question is answered by looking at what the constraint was before, not by looking at what was built. A household that builds a second bathroom because six people share one has answered it. A household that builds a third has not, and the receipt looks identical.

A chain that fails fails at one link, so the question to ask is which of the four links is the binding one in the case at hand, and naming that link is more useful than any verdict on the chain as a whole. The discipline that follows is worth having: something precise and defensible can be said about a specific case without any need to pronounce on whether the whole approach is sound.

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