Expansion vs Recession: Why the Two Are Not Mirror Images

An expansion is output rising, usually for several years together, with employment following it up late and credit growing alongside. A recession is output falling, usually over a much shorter stretch, with employment following it down late as well. The two differ in direction, in length, in speed and in how the damage lands, and those differences matter more than the direction.

Underneath that sits one fact worth fixing in place before anything else. An economy is not a machine that runs forward and then runs backward through the same steps. An economy is a very large number of separate decisions to produce, to hire, to lend and to spend, and those decisions are made by different people at different speeds. When conditions improve, some of those people move quickly and most move slowly. When conditions worsen, almost everybody moves quickly at once. So the upswing and the downswing are built by the same actors behaving in genuinely different ways, and there is no reason at all to expect the two halves to be reflections of each other.

The everyday version carries most of the argument. A tiffin kitchen delivers lunch boxes to office workers. Orders grow, so the kitchen hires one extra cook, waits to see whether the orders hold, then hires another the following season, then buys a second delivery scooter on credit because the finance agent is happy to lend to a business whose order book is filling. The build up takes three years and nobody notices it happening. Then the offices nearby cut their staff and orders drop by a fifth. The two extra cooks and the scooter are not unwound over three years. A kitchen with falling orders cannot carry two extra cooks and a scooter, and everybody in the chain, the owner, the cooks and the finance agent, works that out at the same time. The two cooks and the scooter go in a fortnight. The rise was assembled slowly by many small decisions. The fall was executed quickly by a few large ones. The gap in speed is the asymmetry at issue.

An expansion is defined first on its own terms, then a recession on its own terms, and only then are the two set against each other. Every figure is worked on the Republic of Sankhya, an invented country, and every rate below is recomputed from the underlying levels rather than lifted from a table.

What is an expansion, taken entirely on its own?

An expansion is a stretch of time over which real outputThe quantity of goods and services an economy produces over a period, counted after the effect of price changes has been stripped out, so that a bigger number means more things rather than dearer things. rises period after period. The definition is complete, and it needs nothing except a series of output levels and the ability to compare each one with the one before.

Work it on Sankhya. Output in year 3 was Rs 17,47,200 crore. In year 4 it was Rs 18,52,032 crore, a rise of Rs 1,04,832 crore. Divide Rs 1,04,832 crore by Rs 17,47,200 crore and the answer is 0.06, or 6.00 per cent. In year 5 output was Rs 19,72,414 crore, a further rise of Rs 1,20,382 crore. Divide Rs 1,20,382 crore by Rs 18,52,032 crore and the answer is 0.065, or 6.50 per cent. Two consecutive rises, and the second one larger than the first in both rupees and rate.

Now the mechanism underneath those two rises. The arithmetic is a record of them rather than a cause. An expansion runs on a loop that feeds itself. Somebody spends. Somebody else has a sale. The seller then needs hours worked to meet the sale, so wages go to whoever works those hours, and that person spends. Each turn of the loop is small. The loop turning many times over many months is what an expansion is made of, and it is why an expansion has no obvious moment of beginning that anybody can point to at the time.

Employment moves inside that loop, and it moves late. Sankhya unemployment was 5.45 per cent in year 3, 4.80 per cent in year 4 and 4.20 per cent in year 5. The unemployment rateThe count of people who are looking for work but not working, divided by the labour force rather than by the population. How the labour force itself is defined and measured is a separate subject. fell by 0.65 points and then by 0.60 points, so 1.25 points came off across the two years while output rose 12.89 per cent over the same stretch. A business meets the first part of new demand with the people it already has, and hires only once the new demand has proved it will stay. So employment improved far less than output did, and it improved afterwards rather than alongside. The kitchen waits a season before the second cook.

Credit grows in an expansion, and it grows for reasons on both sides of the transaction at once. A lender looks at a stretch of rising sales and sees borrowers whose repayment looks safer than it did, so it extends more and on easier terms. A borrower looks at the same stretch and sees an investment that now looks worth making, so it asks for more. Neither side is being reckless in any obvious way. Each is responding to evidence, and the evidence is real while it lasts. The Sankhya path fixes output, jobs and prices and does not fix a credit series, so the direction of credit is the only part of it available.

Demand presses against what the economy can actually supply, so prices generally rise faster as an expansion runs on. Sankhya inflationThe rate at which the general level of prices rises over a period. How it is measured, and which index is being quoted, is a separate subject with its own set of choices. ran 5.00 per cent in year 4 and 7.00 per cent in year 5. The pattern is ordinary rather than binding, a tendency rather than a law.

What is a recession, taken entirely on its own?

A recession is a stretch of time over which real output falls. The direction is the only part of the definition everybody agrees on. Where exactly the line falls, and whether it is drawn by a mechanical rule counting consecutive falling quarters or by a considered judgement across many measures at once, is a separate question settled in its own place, and nothing below depends on which of the two answers is preferred. The falls themselves are worked below.

Work them on Sankhya. Output in year 5 was Rs 19,72,414 crore. In year 6 it was Rs 19,52,690 crore, a fall of Rs 19,724 crore. Divide Rs 19,724 crore by Rs 19,72,414 crore and the answer is 0.01, a fall of 1.00 per cent. In year 7 output was Rs 19,03,873 crore, a further fall of Rs 48,817 crore. Divide Rs 48,817 crore by Rs 19,52,690 crore and the answer is 0.025, a fall of 2.50 per cent. Two consecutive falls, and the second larger than the first.

The mechanism is the same loop, running the other way, and it is worth walking it step by step because every step has a person in it. Somebody stops spending. Somebody else has a smaller sale. The seller needs fewer hours worked, wages are not paid to whoever would have worked them, and that person stops spending. Nothing new has been introduced. The identical chain that carried the expansion carries the contraction. The direction of output is, for exactly that reason, the least interesting difference between the two.

Employment follows down, and again it follows late. Sankhya unemployment was 4.20 per cent in year 5, 5.10 per cent in year 6 and 6.90 per cent in year 7, so it rose 0.90 points and then 1.80 points, 2.70 points across the two years. Hold that figure beside the expansion. Output fell 3.48 per cent across the contraction and unemployment rose 2.70 points. Output rose 12.89 per cent across the expansion and unemployment fell 1.25 points. The jobs number responded roughly eight times as strongly per per cent of output on the way down as it did on the way up, and that single comparison is the sharpest of the differences between the two phases. The arithmetic of that ratio is worked out further down.

Credit withdraws in a recession, and it withdraws faster than it was extended. The lender that watched three years of rising sales before easing its terms needs one bad quarter to tighten them. A borrower that spent two years building the case for a new line stops the plan in a week. The lender is not behaving badly. The same evidence-following behaviour is now running against a shorter and more alarming set of evidence, and the speed is the point rather than the direction.

Demand has stopped pressing against supply, so prices usually rise more slowly in a recession. Usually is doing real work in that sentence. On the Sankhya path prices ran 8.00 per cent in year 6 and 6.00 per cent in year 7, so the fastest price rise of the whole stretch landed in the first year output fell. The combination of falling output and fast rising prices has its own name and its own treatment, and it is covered separately.

Now that both are built, where do the two actually part?

Both sides now exist independently, so the comparison can be made without either one leaning on the other. Four criteria are enough, and the direction of output is deliberately listed first so it can be got out of the way.

Direction could be guessed from the words alone. Direction is therefore the least interesting difference between an expansion and a recession. Output rises in one and falls in the other. Everything a reader gains from putting the two side by side sits in the other three rows.

The second criterion is length. An expansion typically runs for years at a stretch. A recession is more often measured in quarters. The tendency describes the usual shape of the record rather than a rule that binds any particular episode. The reason for the tendency is the one already worked above: the upswing is assembled by many small decisions taken at different times, and assembling takes longer than dismantling.

The third criterion is employment, and it splits into two questions that are easy to run together. One question is what employment does. The other is when it does it. In an expansion the unemployment rate falls, late and by little. In a recession it rises, late again but by a great deal more per per cent of output. The lateness is common to both. The size of the response is not.

The fourth criterion is credit, and it is the one a lending desk feels first. Credit is extended gradually in an expansion as evidence accumulates, and it is withdrawn quickly in a recession as evidence turns. The asymmetry in the credit row is not a difference of direction. The asymmetry is a difference of speed, and it shows up in a lender's own accounts long before it shows up in an output series.

Why does an expansion usually add more than a recession takes away?

A reader who has only ever heard the word cycle usually holds a wrong picture of what a cycle does. Slow down here. Take the whole of the Sankhya stretch, years 3 to 7, and lay the four measures out together.

Where these figures come from. Every rupee and every percentage above belongs to the Republic of Sankhya. The path was fixed once, so a level quoted against one year agrees with the same level quoted against that year anywhere else.

Now do the two additions that matter, and do them separately. The expansion: Rs 1,04,832 crore in year 4 plus Rs 1,20,382 crore in year 5 gives Rs 2,25,214 crore added. Check it the other way, Rs 19,72,414 crore less Rs 17,47,200 crore is Rs 2,25,214 crore, and the two routes agree. The contraction: Rs 19,724 crore in year 6 plus Rs 48,817 crore in year 7 gives Rs 68,541 crore taken back. Check it the other way, Rs 19,72,414 crore less Rs 19,03,873 crore is Rs 68,541 crore, and again they agree.

Rs 2,25,214 crore went on and Rs 68,541 crore came off. The expansion added 3.29 times what the contraction removed, and the economy ended the full cycle Rs 1,56,673 crore, or 8.97 per cent, above where it began. Confirm that too: Rs 19,03,873 crore less Rs 17,47,200 crore is Rs 1,56,673 crore, and Rs 1,56,673 crore divided by Rs 17,47,200 crore is 0.0897, or 8.97 per cent when rounded to two decimals.

Two things drive that gap, and only one of them is about length. The first is length itself. Two years of rising against two years of falling is a short and even illustration by choice. A real upswing usually runs several times longer than the downswing that follows it, and that widens the gap further. The second is the size of the annual moves. Plus 6.00 and plus 6.50 per cent are much larger numbers than minus 1.00 and minus 2.50 per cent, and compoundingApplying each period rate to the level the previous period ended at, rather than to the original level, so that changes build on changes. means each of those percentages is applied to a level the previous year already moved.

Here is the household version, and it is the one to hold on to. Somebody starts on Rs 20,000/- a month and gets two raises, then takes a pay cut when the business does badly. The pay cut does not put them back on Rs 20,000/-. The cut puts them somewhere between the raise and the starting point, and the starting point is now well behind them. None of that is optimism about pay. Pay arithmetic runs on a sequence of percentages applied to a moving base, and the same arithmetic is what makes an economy larger at the end of a full cycle than it was at the beginning.

Why does a recession feel sharper than an expansion of the same size?

Suppose for a moment that the arithmetic were exactly symmetric. Suppose an economy added 5.00 per cent over four years and then lost 5.00 per cent over four years, so the rupees on and the rupees off were identical. The two halves would still not feel the same, and there are three reasons, none of which is about the size of the numbers.

The first reason is speed. Output on the Sankhya path fell 2.50 per cent in year 7 alone, having risen 6.50 per cent in year 5. A business responds to a bad quarter faster than it responds to a good one, so the falls are smaller but they arrive concentrated. A good quarter might be luck. A bad quarter threatens payroll, and everybody in the chain reacts to it in the same month.

The second reason is how the employment effect lands, and this is where the numbers do the arguing. Work the response per per cent of output in each direction. Across the expansion output rose 12.89 per cent and unemployment fell 1.25 points, so 1.25 divided by 12.89 is 0.10 points of unemployment per per cent of output. Across the contraction output fell 3.48 per cent and unemployment rose 2.70 points, so 2.70 divided by 3.48 is 0.78 points per per cent of output. The unemployment rate moved roughly eight times as far per per cent of output on the way down as it did on the way up. The ratio is the arithmetic behind a phrase most people only ever meet as a feeling.

The third reason is who carries it. Gains in an expansion are spread thinly across many people: shorter job searches, an extra shift, a small raise. Losses in a recession are concentrated: a small number of people lose an entire income at once while everybody else keeps theirs. An economy that loses 2.50 per cent of output has not made everybody 2.50 per cent poorer. The fall has left most people roughly where they were and taken almost everything from a few. A smaller number in the aggregate can therefore produce a much louder result in a country, and the concentration of the loss is the single most important thing to remember when reading a headline that reports a fall of a per cent or two.

There is a fourth item, and it belongs here because it is the lesson the next stretch of these notes builds on. Sankhya output turns up in year 8, growing 5.00 per cent, and the unemployment rate keeps rising anyway, from 6.90 to 7.40 per cent. The jobs number had not turned a full year after output had. A reader watching the unemployment rate alone would have declared the recovery absent while it was already a year old. Employment does not merely follow output. Employment follows late enough to mislead an observer about which phase the economy is in.

What do prices do in each, and where does that question belong?

The tendency is easy to state and the exceptions are where the interest lies. Demand presses on what the economy can supply as an expansion matures, so prices usually rise faster. In a recession that pressure has gone, so prices usually rise more slowly. Sankhya shows the first half of that tendency cleanly: prices ran 5.00 per cent in year 4 and 7.00 per cent in year 5, firming as the expansion ran on.

Sankhya then refuses to be tidy. Look at what it does next. Prices ran 8.00 per cent in year 6, the fastest reading of the whole stretch, in the first year output actually fell. Only in year 7 do they ease to 6.00 per cent, and by year 8 to 3.00 per cent. So on this path the price reading was at its worst exactly when the output reading first turned bad. Falling output beside a fast rising price reading has earned its own name and its own treatment, set out separately.

Output direction and price direction are related but not locked together. A reader who infers the phase from the price reading alone will be wrong precisely in the cases that matter most. How prices are measured, which index carries which basket, and why two price measures can disagree about the same month are all separate subjects with their own machinery.

Why does a lending book behave asymmetrically through a cycle?

Everything above is about an economy. The asymmetry stops being an abstraction as soon as a single lender inside that economy is examined.

Take an invented lender inside Sankhya with a book of Rs 1,000 crore, and give it a margin of 2.50 per cent a year on that book, so it earns Rs 25 crore in a year. Across the two expansion years it earns Rs 25 crore and Rs 25 crore, a total of Rs 50 crore. Now the contraction arrives. Suppose 5.00 per cent of the book, Rs 50 crore, stops paying. Suppose Rs 15 crore of that is eventually recovered, leaving Rs 35 crore written offRemoving an amount a lender no longer expects to collect from the value of its loans, so that the loss is recognised in its accounts in that period.. The Rs 35 crore lands in a single year. Rs 35 crore divided by Rs 25 crore is 1.40, so one year of losses cancels one year and five months of margin, and the lender does not get those months back when the economy recovers.

Notice what has and has not happened there. The lender did nothing different in the two phases. Its rate did not change, its book did not change size, and the arithmetic in its spreadsheet was symmetric all the way through. The timing is what differs: margin arrives in thin monthly slices while a write-off arrives whole. A lending desk therefore watches the jobs number and the credit growth number even though its own contract is written in rupees of interest, and a downturn shows up in a lender's accounts as a step rather than a slope.

An analyst reading such a lender applies the same asymmetry to the outlook rather than to the history. The expansion never tested the book, so two years of clean recoveries say very little about how it behaves in a contraction. A household reading its own position can use exactly the same thought: an income that has been reliable through a long expansion has not been tested against a contraction, and the honest response is to notice that, not to assume it either way.

The Republic of Sankhya, the tiffin kitchen, the salary of Rs 20,000/- a month and the lender with a book of Rs 1,000 crore are invented.
Educational material. Not advice on any investment, tax, budget or market position.