Sectors in Macro Analysis: How the Economy Is Divided

An economy gets divided three common ways: by what is produced, by who buys the output, and by how completely the activity is recorded. Each division answers a different question. Dividing matters because one growth rate is an average over parts that move at different times and by different amounts, and the average itself carries none of that.

Underneath the answer sits a single arithmetic fact that most readers already accept and few readers use. An average is built by throwing away the differences between the things being averaged. Discarding the differences is not a defect of averaging, it is the whole job of averaging. So a country that reports one output figure and one growth rate has reported something complete and correct. The next question anybody asks is who exactly had the bad year, and the report discarded that answer on the way to being correct.

The pieces needed are already in place. Output came earlier, and so did the output gapThe difference between what an economy actually produced and what it could have produced without straining, expressed as a percentage of the second. The output gap is built and worked through separately.. So did inflation and the way it is measured. The four phases of a cycle came at the start of this stretch of reading, and so did the path that the Republic of Sankhya, an invented country used as a working model, walks through them. The aggregate a reader already knows how to read now gets cut open.

Why divide an economy at all, when the growth rate already says what happened?

Start with a household rather than a country. A married couple keep one joint account. In a particular year the account balance falls by one per cent, and read on its own that is a mild year. Open the two salary slips inside it and the picture changes completely: one salary was cut by a third when the shop the husband works at reduced hours, and the other rose by ten per cent because the wife was promoted. The joint balance is exactly right and it is also a description of nothing that happened to either person. One of them had a very bad year.

An economy is that joint account with a few hundred million people inside it. The aggregate is never wrong and it is always incomplete, and dividing it is how the information the average threw away is recovered. There is a common overcorrection to guard against here. The headline number is not misleading, badly compiled or unworthy of reading. The headline number is a good answer to the question it was built for. The headline number simply cannot be interrogated for the answer to a different question, and no amount of staring at it will make it give one up.

There is a second reason to divide, and it is the reason the subject needs a treatment of its own rather than a paragraph somewhere else. Parts of an economy do not move at the same time. The parts do not merely move by different amounts in the same year. They take their turn in different years. A country can be one year into a downturn in one activity and not yet in it at all in another. Once that is understood, a single number for the whole country stops being a summary of a situation and becomes a snapshot of several different situations that happen to be added together.

What does each of the three divisions actually reveal?

The three divisions are genuinely different tools, and confusing them is the most common way a reader gets a sector conversation wrong. Take them one at a time.

The first cut is by what is produced. Growing things, making things, and doing things for people. The producing cut answers what a country does for a living, and it changes shape over decades rather than over quarters. The producing cut also builds its arithmetic up from the value each producer adds rather than from what anybody spent, so a reader meets it in the company of gross value addedOutput measured as the value each producer adds, before taxes on products are added and subsidies taken off. Gross value added is the production side of the accounts, and it is defined in full where the production measures are built. rather than in the company of spending.

The second cut is by who buys the output. Households, businesses, government, and buyers abroad. The buying cut answers a different and sharper question: whose decision has to change before output changes. A downturn is not an event that happens to production, it is a set of decisions made differently by people who buy, so the buying cut is the one most useful in a downturn. To know why output fell, the producing cut shows where the fall appeared and the buying cut shows who caused it.

The third cut is by how completely the activity is recorded. Some activity sits inside arrangements that a statistical system reaches easily, and some sits in arrangements it reaches less completely. The recording cut is a real and important division, it is not the same thing as either of the other two, and it is worked through with the growth material. An economy does not divide only two ways.

Each of the three is a different question, so the right division is chosen by what is being asked, and no one of them is the correct division. A reader who wants to know what an economy does uses the first. A reader who wants to know why a year went badly uses the second. A reader who wants to know how much the measurement is even seeing uses the third. Asking which is the true division of an economy is like asking whether a map of roads or a map of rivers is the true map of a district.

How Business Cycles Move Sectors at Different Times: what decides the order?

Almost everybody meets the ordering of sectors first as a list to memorise, and a list is exactly the wrong way to hold it. There is one mechanism underneath, and once that mechanism is held the ordering for a sector nobody has mentioned can mostly be worked out from scratch.

Think about what actually happens in a household when the news turns bad and the earner is not sure about next year. Nothing dramatic happens on the first day. Nobody stops eating, nobody cancels the electricity, nobody hands back the house. The new scooter waits. The kitchen renovation waits. The old phone still works, so the replacement waits. The first thing that happens when confidence falls is not a cancellation, it is a postponement, and only purchases that can be postponed are available to be postponed.

Now scale the household up to a business, which does the same thing with bigger amounts. A factory that is unsure about next year has orders to fill this month, so it does not stop buying raw material. The factory stops the new machine. The machine was going to replace one that still runs. Nothing breaks if the decision waits a year, so the decision waits a year, and it waits within weeks of confidence turning rather than within years.

Postponement gives the rule that everything else turns on. Timing follows postponability: a sector is reached early when its output is bought on a decision that can be put off, and late when its output is bought on a commitment already made or bought regardless of the weather. Three positions come out of that one idea.

A sector is reached early when the buyer has full discretion over when to buy and no penalty for waiting. Equipment that replaces working equipment sits here. So does anything bought out of the part of a household budget that is genuinely optional.

A sector is reached late when the buying decision was made a year or two ago and cannot now be unmade. The money on a signed multi-year contract was committed before anybody was worried, and walking away would cost more than finishing, so a builder keeps building through the first year of a downturn. A contracted sector has an order backlogWork already contracted and paid for or committed to. A producer keeps delivering it regardless of what has happened to demand since the order was signed. standing between it and the downturn, and the backlog runs out at some point rather than never. When it runs out, the sector gets its downturn, one or two years after everybody else got theirs.

A sector is barely reached at all when the purchase is not a decision in any meaningful sense. Nobody sits down and decides whether to buy food this week. The quantity moves a little, the mix inside it moves more, and the total barely notices what the rest of the economy is doing.

Notice the omission in the rule. Being reached late is not the same as being spared, and being reached early is not the same as being damaged, so the rule says nothing at all about which sector is a better place to be. A sector reached early may recover early too, for exactly the same reason: a postponed purchase was postponed rather than abandoned, so it is still waiting to happen and it arrives in a rush when confidence returns.

Watching one contraction reach three sectors in three different years.

The Republic of Sankhya contracted 1.00 per cent in year 6 and a further 2.50 per cent in year 7. Both figures are fixed by the Sankhya path built earlier, and the sector split below has to agree with them. Three invented sectors divide that same output, chosen so that they add back to the published total in every single year. Three sectors can differ enormously while adding up exactly.

Meet the three, all invented. Yantra makes replacement equipment for other businesses, so almost every rupee it earns depends on somebody choosing not to wait. Setubandha builds infrastructure under contracts signed years before the work is delivered. Annapurti processes and distributes weekly staples. Their year 5 shares of Sankhya output are 26.00 per cent, 30.00 per cent and 44.00 per cent respectively.

The bottom two rows are the check that makes everything above them trustworthy, so read them first. In year 6, Rs 4,71,802 crore plus Rs 6,00,008 crore plus Rs 8,80,880 crore comes to Rs 19,52,690 crore, the published Sankhya figure to the last rupee crore. In year 7 the same addition gives Rs 19,03,873 crore, again exactly the published figure. One aggregate contains all three behaviours at once, and the fact that the parts add up is precisely why the aggregate could never have shown any of them.

Now read the sector rows and watch the years. Yantra took its whole fall in year 6, minus 8.00 per cent, eight times the size of the aggregate move in the same year. In year 7 the aggregate fell harder than it had in year 6, and Yantra fell only 0.43 per cent. The purchases that could be postponed had already been postponed. Setubandha did the opposite. Setubandha grew 1.40 per cent in year 6, delivering work signed before anybody was worried, and then fell 10.00 per cent in year 7 when the backlog ran out and no replacement orders had been signed. Annapurti did the same thing in both years, growing 1.50 per cent while the country around it shrank twice.

One fact here is worth carrying away, and a list would never give it. The measurement runs from the year 5 peak to the year 7 troughThe lowest point output reaches before it starts rising again. The four turning points of a cycle are named separately and are taken as read here. and Yantra is down 8.40 per cent while Setubandha is down 8.74 per cent. Over two years the two of them had almost the same downturn, 0.34 percentage points apart. The two sectors were separated not by how much they fell but by when, and a reader looking only at year 6 would have concluded that one was in trouble and the other was untouched. Both statements would have been true in year 6 and both would have been useless by year 7.

Why does the same growth rate mean different things to different sectors?

Setubandha grew 1.40 per cent in Sankhya year 6. Held on its own, that number tells almost nothing, and the reason is worth being precise about.

In year 6 the country around Setubandha shrank 1.00 per cent, so growing 1.40 per cent put the sector 2.40 percentage points ahead of everything else. In year 5 the same country grew 6.50 per cent, and the identical sector rate of 1.40 per cent would have put it 5.10 percentage points behind. Same rate, two opposite readings, and nothing about the sector changed between them. A sector growth rate is a comparison waiting for its second term, so a sector number without the aggregate beside it cannot be interpreted at all.

A salary rise of six per cent is different news depending on what happened to prices, and the reason is the same one, already met in real termsA figure after the effect of price changes has been taken out, measuring quantity rather than the rupees the quantity happened to cost. Real terms is covered in full separately.. The number on its own is not the fact. The number plus the figure it must be read against is the fact.

There is one more layer, and it is the reason a big sector and a small sector cannot be compared by their rates alone. A sector contributes to the aggregate in proportion to its size, so its rate is scaled by its weightThe share of the total that a part represents. The share decides how much that part pulls on an average, and a part with a small weight can move a long way and barely move the total. before it reaches the total. Annapurti is 44.00 per cent of Sankhya output, so its steady 1.50 per cent adds 0.66 percentage points to the aggregate on its own. Yantra is 26.00 per cent, so its dramatic minus 8.00 per cent subtracts 2.08 points. The two contributions, plus Setubandha's 0.42, are exactly the minus 1.00 per cent the country reported.

What does an analyst actually do with a sector reading?

The most useful move an experienced analyst makes with a sector label is to distrust it for one specific reason, and it follows directly from the postponability rule rather than from any list.

A sector label describes the buyer the sector is usually assumed to have. A label does not describe the buyer a particular company inside the sector actually has. So the question is never what sector a company is filed under, it is what the company's customer is deciding when the customer buys from it. A firm that makes stainless steel vessels for staple food processors is filed alongside staples. Its customers buy vessels the way anybody buys equipment, and they postpone the purchase the moment they are unsure. The vessel maker behaves like the equipment, not like the label, and the label will be wrong about it in exactly the year it matters most.

The same move works in the other direction and is easier to miss. A company inside a sector that everybody expects to be reached early may sell almost entirely under maintenance agreements that run for five years. Its revenue is committed, not discretionary, and it will be reached late even though its neighbours were reached first.

The practical version of the move fits in three questions, and none of them requires a forecast. Who signs the cheque, a household or a business or a government body. Can that signer wait a year without a penalty. And was the decision already made, under an agreement that predates the current mood. A credit officer sizing exposure, an equity analyst deciding which quarter to expect the revenue hit in, and a household deciding how secure one earner's work is are all asking that same set of three questions about different things.

Where is a real economy actually divided, and who publishes it?

A mechanism survives a decade, and a magnitude belongs to one particular year, one particular definition and whoever compiled it. So everything above was worked on an invented country. The magnitudes for a real economy exist, they are published, and they are published by named bodies whose documents are open to any reader.

Does any of this say which sector is the better one to be in?

No. Sector reading gets turned into something it is not at exactly this point, so the answer is flat rather than hedged.

Everything above is a description of when a cycle arrives somewhere and why it arrives then. Which sector is preferable, how the three rank against each other, and whether a phase of a cycle calls for holding anything are different questions altogether. Those are questions about positioning, they are answered in a different subject area entirely, and they need inputs that never appear above: what a thing costs relative to what it produces, how long it is intended to be held, and what follows if the judgement is wrong.

The temptation to rank the three is strong, and it fails for a reason worth seeing. Yantra fell 8.00 per cent in year 6, so it looks like the sector to avoid until the next year arrives. Yantra fell only 0.43 per cent in year 7, and Setubandha, the sector that looked untouched, fell 10.00 per cent. Being reached first is not the same as being harmed most, and a reader who converted timing into preference would have been wrong within twelve months on invented figures that were built to be internally consistent. On real ones, with prices attached, it would be worse.

The Republic of Sankhya, Yantra, Setubandha and Annapurti are invented.
Educational material. Not advice on any investment, tax, budget or market position.