How to Read the Union Budget for Investors

Read a budget in a fixed order: the three totals first, then the gap as a share of output, then the composition of each side, then the individual proposals, and the speech last. The order is the method. Every figure after the first three is a share of one of them, and a reader who starts anywhere else is holding a number with no denominator.

The seven steps below are an order of operations and nothing more. No step below pauses over what a deficit is, why a government sorts its spending the way it does, or where its receipts originate; all three are covered separately, and a step that begins teaching one of them has stopped being a step. Three things are taken for granted throughout and never repeated: that a government acts on the economy through spending and taxing; that the gap between those two has three named versions which differ in what they leave out; and what kind of thing a budget document is, what its parts carry, and how one figure inside it can be a different kind of figure from the one beside it. The order leans on all three, so any of them that feels thin is worth going back for first.

In what order should a budget document be read?

Seven steps, and the order is not a matter of taste. Consider a hospital bill, or a caterer's quote for a wedding. The eye of most readers lands first on the single biggest line, the one that stings. A person who reads well does the opposite, finding the total at the foot of the bill, then the amount already paid, then the balance still due, and only after those three going looking for what the Rs 40,000/- line was for. By then Rs 40,000/- has a size: it is a fifth of the bill, or it is nine tenths of it, and those are two completely different kinds of news. Read first, the same Rs 40,000/- is only a large number and a feeling.

A budget document is that bill at national scale, and the trap in it is exactly the same trap wearing better clothes. The steps below are ordered so that nothing is ever read before the figure that gives it a size. Each step is an action, each action produces something written down, and the next step uses only what the last one produced. No step calls for a decision about the government, and none explains why the machinery works. The machinery is built elsewhere, and none of it is needed to take a step.

What are the first three figures taken, and why those?

Step one: write down total spending, write down total receipts, and write down the difference between them. Three figures, each with its unit attached, and nothing interpreted. On the document of the Republic of Sankhya, an invented state used throughout, that is Rs 4,00,000 crore going out, Rs 3,00,000 crore coming in, and a gap of Rs 1,00,000 crore. The three totals go first for a single reason: nearly everything else printed further down turns out to be a fraction of one of them.

Step one does very little. Step one does not name the gap, does not say which of the three versions of the gap this is, and does not say a word about what any of it means. Whether that Rs 1,00,000 crore should be called a fiscal deficit, and how it differs from the two narrower versions that leave out interest or leave out capital, was set out in the comparison of the three deficits and is not needed here. Step one needs only the subtraction, and a subtraction does not require the name of its answer. The unit travels with every one of the three. A figure in Rs crore sitting next to a figure in croreAn Indian unit of counting equal to ten million, written with the digits grouped as 1,00,00,000. Rs crore therefore means the amount in rupees divided by that, and the unit has to travel with the number or the number means nothing. without its unit is the commonest way a reading goes wrong in the first thirty seconds.

Why does the gap go against the output measure before anything else?

Step two: divide the gap by the output measure and write the share to two decimals. On the Sankhya document that is Rs 1,00,000 crore against published output of Rs 17,47,200 crore, and the division comes to 5.72 per cent. The division is the whole of step two, and its output is one number.

Why does this come second rather than later? Because a rupee amount at national scale is unreadable to a human being. Nobody has an instinct for whether Rs 1,00,000 crore is large. But everybody has an instinct for a share, and the output measureThe published total value of what an economy produced over a period. It is built and reported separately, and here it is simply borrowed as the denominator that the rest of the document is read against. is the one denominatorThe number a figure is divided by. Change it and the same figure on top produces a completely different answer, which is why a share is meaningless until the denominator beside it is stated. that stays the same size in every part of the document. Fixed at this point, it gives every later figure something to be put against. Left out, it leaves the rest of the reading a matter of comparing rupee amounts to each other with no sense of scale at all.

Step two produces a share, and nothing else. No step in the order produces a judgement about whether 5.72 per cent is a lot. Why borrowing happens and where the money is raised are left out too. Sorting the gap into its named versions belongs to the comparison of the fiscal, revenue and primary deficit, and adds nothing to a step whose job is one division.

How is the spending side split without the reader deciding anything?

Step three: read the labels the document has already put on its own spending lines, group the lines by label, and add each group up. Two subtotals come out. On the Sankhya document the lines labelled revenue add to Rs 3,30,000 crore and the lines labelled capital add to Rs 70,000 crore. Against the Rs 4,00,000 crore total those two are 82.50 per cent and 17.50 per cent, and step three ends there.

The important word in that step is read, not decide. A budget document states which of its lines are which. Nothing is being classified at this step; a classification somebody else already applied is being transcribed, and then added up. The difference between reading and deciding matters more than it sounds. A reader who starts deciding which spending counts as which has wandered into a genuinely difficult subject at exactly the point where arithmetic is what is needed. The meaning of those two labels, the argument over the split between them, and the places where the classification gets hard at the margin are opened up in the treatment of government spending as revenue against capital. Step three borrows none of it, and it can afford not to because the document has done the labelling and the reader is only adding.

So the household version. An electricity bill arrives with a fixed charge, an energy charge and a late fee already itemised, and nobody sits down to rethink which is which. The lines carrying the same tag get added together. Deciding halfway through that the late fee was really an energy charge would create an interesting argument and destroy any comparison between this bill and last month's. Capital spendingThe lines a budget document files under building or acquiring something that lasts. What the label covers, and why it is contested at the edges, is a separate subject and is not settled here. and its counterpart are the same situation at national scale: the tag is already on the line, and step three only sums by tag.

What does the receipt side look like once it is written as shares?

Step four: read the receipt lines exactly as the document reports them, turn each into a share of the receipts total, and note which kinds the total leans on. On the Sankhya document, off a receipts total of Rs 3,00,000 crore, direct tax is Rs 1,20,000 crore at 40.00 per cent, indirect taxA tax that attaches to a transaction rather than to a person, so the one who hands it over and the one who bears it can be different people. How the two kinds compare is treated on its own elsewhere. is Rs 1,50,000 crore at 50.00 per cent, and non-tax receiptsReceipts raised by some route other than taxation, such as charges for a service or returns on holdings. The full set of routes is built as its own subject. are Rs 30,000 crore at 10.00 per cent. Tax of one kind or the other is therefore 90.00 per cent of the total. Three shares, and they are the whole output of step four.

All step four needs is that receipts arrive in kinds and that two of those kinds are taxes. Nothing else. Step four does not require working out why one kind is bigger, what happens to any of them when the economy turns, or which is the fairer way to raise money. Where the receipt side comes from in full is built as its own subject, and the contrast between the two kinds of tax is treated separately again. Step four takes nothing from either and simply writes down three shares that add to a hundred.

How is an announcement separated from an authorisation while reading?

Step five takes the proposals one at a time and puts each into one of two columns before any of them is read on its merits. One column is for a proposal with an amount attached to a line in this document. The other is for a proposal that has been stated and has no amount attached to anything here. The test for which is which was set out when the document itself was taken apart, and it is applied here rather than re-argued. The output is one list in two columns, and the sorting happens before the reading rather than after it.

On the Sankhya document, three proposals carry an amount: a coastal roads programme at Rs 2,800 crore, rural water works at Rs 4,200 crore and a teacher recruitment drive at Rs 1,500 crore. Three others do not: a commission to review the tax on transactions, an intention to bring three port trusts under a single authority, and a review of fees on state services. Six proposals, three in each column, and not one word yet about whether any of them is a good idea.

Why sort before reading rather than after? Because both kinds are written in the same voice, and voice turns out to be a terrible guide. Consider the notice a building association pins up in the lobby: the lift repair has been contracted at Rs 1,80,000/-, and the parking arrangement will be reorganised shortly. One of those has money behind it and a receipt somewhere. The other is somebody's intention, quite possibly a sincere one. A sentence never announces which of its halves is backed by an amount, so a single reading of that notice leaves a resident with two settled facts. Ticking each item as it goes past is the only cheap fix, and it has to happen on the way through. By the end the two have already fused into a single memory.

What stops two figures in the document from being compared?

Step six checks what kind each figure is before putting any two of them side by side, and where the kinds differ, the word refused is written and the reading moves on. The output is every comparison attempted, each marked either like for like or refused. A budget document carries figures of several kinds, it labels which is which, and the labels are there precisely because the figures are not interchangeable. The kinds of figure a budget carries, and the way the document signals them, were covered when the document itself was taken apart. Step six only reads the label and obeys it.

Three attempts on the Sankhya document show the shape of it. Coastal roads at Rs 2,800 crore against rural water works at Rs 4,200 crore: both are authorised amounts in the same column of the same document, so the comparison runs, and the second is roughly one and a half times the first. Coastal roads at Rs 2,800 crore against the same programme's figure in the column for the year now running: refused. The document labels those as two different kinds, and step six does not read past a label it has not checked. The capital subtotal against the port trusts announcement: refused. One side is an amount, and the other side has no amount anywhere in the document. Two of three attempted comparisons produce nothing, and writing refused is a real output rather than a failure to find one.

What would have to change for this reading to change?

Step seven writes down two or three specific things that, if they happened, would move an output already recorded. Each condition names a figure, a direction and a size, and every one of them is written now, before the year has run and before anybody knows how it turned out. Writing the conditions in advance is what keeps a reading honest. A test written after the result is known can always be bent until it fits the result.

Anybody who has ever said afterwards that they knew all along knows why this matters. A condition set down in advance can be checked by somebody else, and it can embarrass its author. A view formed afterwards cannot. Three for the Sankhya document follow, and not one of them is a prediction.

Every figure in this table is recomputed from the Sankhya account rather than carried over: Rs 4,00,000 crore less Rs 2,85,000 crore is Rs 1,15,000 crore, and that against output of Rs 17,47,200 crore is 6.58 per cent. Rs 3,30,000 crore plus Rs 55,000 crore is Rs 3,85,000 crore, and the two shares of it are 85.71 and 14.29 per cent, adding back to a hundred.

When does the reading stop?

The stopping point is a step, not something that arrives with fatigue. The reading stops when the outputs exist: three totals, the gap as a share, two spending subtotals, three receipt shares, the proposal list in two columns, every comparison marked or refused, and the named conditions. The count settles it: with all seven written down the reading is finished, and with one missing it is not, however long it has gone on.

The reading does not stop with a view about the budget, and that is worth being blunt about rather than apologetic. The seven steps do not produce a verdict, they produce material, and the two are different things that people routinely confuse. Nowhere in the seven steps is there an instruction to decide whether the gap is too wide, whether the split between the two kinds of spending is the right one, or whether any proposal should have been made. All three are judgements, they need arguments this reading never touches, and a reader who reaches the end holding one has smuggled it in from outside rather than derived it here.

What does the sequence look like run end to end on one document?

Here is the whole thing on the Sankhya document, from the first subtraction to the stopping point. Read down the outputs column and notice how small it is. Six figures, a two-column list and three conditions is the entire product of a full reading, and there is not a sentence of opinion in it.

Where these figures come from. Every rupee amount worked through here belongs to the Republic of Sankhya, a state invented for this sequence and locked so that its account balances the same way wherever it is used. The institutions listed lower down are real, and they appear here purely as addresses; whatever they publish has to be looked up rather than recalled.

One thing that table makes visible is the reason the order is fixed. Every figure below the first three is a share of one of the first three. The two spending subtotals are shares of total spending. The three receipt shares are shares of total receipts. The gap as a share is taken against the output measure. Take the first three away and every remaining number loses its size instantly. A number with no denominator is what the answer at the top meant, and that is the whole justification for the order.

What does a reader lose by starting with the speech?

Now the failure, and it is the commonest one there is because it is what the document itself invites. The speech is written to be read first. The speech is in plain language, it has a story in it, and the numbers inside it are chosen precisely because they sound like a lot. So a reader opens there, lands on a scheme figure, and forms an impression in about four seconds.

Take the coastal roads programme at Rs 2,800 crore. Read first, that figure is enormous and completely sizeless at the same time. Nothing has been written down yet to divide it by. Read last, after six steps, the same Rs 2,800 crore is 4.00 per cent of the capital subtotal, 0.70 per cent of total spending, 2.80 per cent of the gap and 0.16 per cent of the output measure. The number never changed; what changed is whether anything existed to give it a size. And note that even at the end there are four different answers, so a reader has to say which denominator they used or they have not really said anything.

What changes when the reading is done for one company rather than the whole economy?

Almost nothing changes, and that is the useful finding. An analyst covering a single business, a lender sizing a borrower's exposure to public spending, or an investor trying to work out whether a sector is about to be busier all run the same seven steps in the same order. The totals are the same totals. The gap goes against the same denominator. The tags on the spending lines are the same tags.

One step gains something, and only one. At step four, alongside the receipt shares, the reader adds a question: which of these receipt kinds bear on the sector I am looking at, and through what? A cement supplier and a consumer lender are exposed to completely different lines, and writing that question down at step four is what stops a reader carrying a general impression of the document into a specific decision. The question is added at step four, and it is asked rather than answered. Answering it needs work this reading does not do.

A careful practitioner, holding the seven outputs, still refuses to skip straight to the sixth. The conditions from step seven are kept somewhere they will be seen again, and in a few months those conditions become the only honest way to check whether the reading held up. Everything else about how a change in government borrowing travels outward into markets, and what happens to the cost of money when a government borrows more, is a separate subject and stays there.

The Republic of Sankhya, its receipts, its spending, its borrowing and every proposal sorted above are invented.
Educational material. Not advice on any investment, tax, budget or market position.