Employment Growth vs Economic Growth: Two Counts

Economic growth counts output. Employment growth counts people at work. Output rises when there are more workers or when each worker produces more, so the two counts can climb together, move apart, or point in opposite directions. An economy can grow five per cent in a year while unemployment rises, and that is ordinary arithmetic rather than a contradiction.

Underneath that sits a fact about counting that is easy to miss. Nobody anywhere measures how well an economy is doing. Measurement produces a set of separate quantities, each collected by different people using different methods for different purposes, and every one of them answers its own narrow question and no other. Output is one such quantity. The number of people at work is another. Output and the number of people at work are related. Output has to be produced by somebody. But a relationship is not an equivalence, and the moment one count is treated as a stand-in for the other the reading has become a guess.

The everyday version is worth holding on to for the rest of this guide. A tailoring workshop takes in Rs 4,00,000/- of stitching work this year against Rs 3,60,000/- last year. Business is up by a ninth. A different question follows: how many people are at the machines? The answer might be nine, the same nine as last year, each one working a little faster because a better cutting table arrived. Or it might be ten, the tenth hired to keep up. The takings figure was never counting people, so it cannot say which. Two questions that sound like one question have two separate answers, and the whole of this guide is the consequence of that.

Economic growth is built first, completely, without a single mention of jobs. Employment growth is built next, completely, without borrowing anything from the first part. Only after both stand on their own are they set against each other. A comparison made before either side is properly defined teaches the difference without teaching the things being differenced. Everything is worked on the Republic of Sankhya, an invented country, and every rate below is computed from the underlying counts rather than quoted from a table.

What exactly does economic growth measure?

Economic growthThe change from one period to the next in the volume of goods and services an economy produces, measured after the effect of price changes has been stripped out. is the change in the volume of goods and services an economy produces, from one period to the next, with price changes taken out. The definition ends there. The calculation needs one series of output levels and the ability to subtract and divide, and it produces one number.

The Sankhya figures make it concrete. Real output in year 1 was Rs 16,00,000 crore. In year 2 it was Rs 16,80,000 crore. The rise is Rs 80,000 crore, and Rs 80,000 crore divided by Rs 16,00,000 crore is 0.05, or 5.00 per cent. Year 3 output was Rs 17,47,200 crore, a rise of Rs 67,200 crore on the year 2 base, and Rs 67,200 crore divided by Rs 16,80,000 crore is 0.04, or 4.00 per cent. Either division redone lands on the same figure.

Notice what that calculation never asked at any point: who produced the output, how many of them there were, whether the number of them changed, or whether any of them were paid. None of that enters the arithmetic, and none of it could, because the only inputs were two output levels. The silence is not a shortcoming that better data would fix. The measure is doing exactly what it was built to do. A production count counts production. Knowing about people means counting people, and that is a different collection entirely.

A farm harvested 100 quintals last season and 110 this season. Output growth is 10.00 per cent, computed from the harvest and nothing else. Whether the extra ten quintals came from two extra hands, from a borewell that arrived in March, or from a season with better rain is invisible in that 10.00 per cent, and every one of those three causes has completely different consequences for the people in that village. The growth figure is honest and useful and it will not say which of the three happened.

What does employment growth count, and how is the unemployment rate different?

Set the output figure completely aside now. The count that follows is built from scratch and borrows nothing from the one above.

Employment growthThe change from one period to the next in the number of people doing work, however that work is defined by the survey doing the counting. is the change in the number of people doing work. Sankhya had 8.00 crore people at work in year 1, 8.00 crore in year 2, and 8.32 crore in year 3. Employment growth was nil in year 2: 8.00 crore did not change. In year 3, 0.32 crore divided by 8.00 crore is 0.04, or 4.00 per cent.

The soft edge in that count is not the arithmetic but the word work. Every statistical system has to define work, and none of them can define it cleanly. A woman who runs a tea stall for four hours a day and looks after two children for the rest of it is doing work by any human standard, and whether she is counted depends entirely on the question the survey asked and the reference period it asked about. A man who helped on the household plot last week without being paid may or may not appear. Somebody who worked two days out of thirty may or may not appear. None of this is an oversight. The decisions are unavoidable, and different decisions produce different counts of the same reality. A definition change can move an employment series without one person changing what they do.

Now the second measure, and it is genuinely a different object rather than the same one restated. The labour forceEverybody who is either working or actively looking for work. Somebody who is neither working nor looking, such as a full time student, is outside it entirely. is everybody who is either working or actively looking for work. Sankhya had 8.40 crore in the labour force in year 1, 8.60 crore in year 2, and 8.80 crore in year 3. Anybody neither working nor looking is outside it altogether. The unemployment rateThe number of people looking for work but not working, divided by the labour force rather than by the population. Both parts of that fraction move on their own. is the number of people looking but not working, divided by the labour force.

Work all three years. In year 1, 8.40 crore in the labour force less 8.00 crore at work leaves 0.40 crore looking, and 0.40 divided by 8.40 is 0.047619, or 4.76 per cent. In year 2, 8.60 crore less 8.00 crore leaves 0.60 crore, and 0.60 divided by 8.60 is 0.069767, or 6.98 per cent. In year 3, 8.80 crore less 8.32 crore leaves 0.48 crore, and 0.48 divided by 8.80 is 0.054545, or 5.45 per cent.

One more measure belongs here because it is the one that catches what the rate hides. The participation rateThe share of the working age population that is in the labour force, meaning working or looking for work. Participation falls when people stop looking, whatever the reason. is the labour force as a share of the working age population. Sankhya has 13.20 crore people of working age, so participation ran 8.40 on 13.20, or 63.64 per cent, then 8.60 on 13.20, or 65.15 per cent, then 8.80 on 13.20, or 66.67 per cent. Employment growth counts a number of people, and the unemployment rate is a fraction with a denominator of its own, so the two can move in ways that have nothing to do with each other.

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

Both are on the table now, each defined without reference to the other, so a comparison is finally honest. Four differences separate them, and they are worth taking one at a time rather than as a blur.

What gets counted is the first and the largest. Economic growth counts a volume of production, expressed in money after price effects are removed. Employment growth counts human beings. Production and human beings are not two views of one thing; they are two things. What a rise means for a household is the second. A rise in output means more was produced. Whether any of it reached the household in question is a separate matter. A rise in employment means somebody who was not working is now working. For that household it is a direct and immediate change in what arrives at the end of the month. How fast the count responds is the third. Output can rise the moment a machine runs longer or a shift is extended, before anyone is hired, and it can fall the moment orders dry up, before anyone is let go. Employers hold on to trained people through a soft patch and wait for proof before hiring into a good one, so the employment count moves later in both directions. How contested the definition is is the fourth. Output has a tightly specified method with published rules. Work has an edge that is genuinely fuzzy, and reasonable definitions disagree.

Put those four together and a pattern emerges: the count that moves first is the one defined most tightly, and the count that matters most directly to a household is the one defined most loosely. That combination is exactly why the two series get confused. The fast, clean number arrives first and gets reported, and the slow, messy number that people actually feel arrives later and gets reported less.

Why do output and work disagree so often?

The mechanism behind every disagreement between the two counts is a single line of arithmetic. Output is the number of people at work multiplied by what each one produces. Written out, that becomes an identity that has to hold, in every economy, in every year, with no exceptions and no assumptions: one plus output growth equals one plus employment growth, multiplied by one plus growth in output per worker.

The everyday reading of that identity is the useful one. Output can only rise in two ways. Either more people are producing, or the same people are producing more each. Any rise in output is some mixture of those two, and the mixture is what decides whether a growth year is also a hiring year. The two counts agree only when output per worker is flat, and any gain in output per worker drives a wedgeThe gap opened between two figures by a third one sitting in between them. Here it is the gap between output growth and employment growth, and its size is the growth in output per worker. between them exactly the size of that gain.

Sankhya makes this visible because in each of the two years one of the two terms is exactly nil. In year 2, output per worker went from Rs 2,00,000/- to Rs 2,10,000/-, a rise of Rs 10,000/- on Rs 2,00,000/-, or 5.00 per cent. The number at work did not change at all. So the whole of the 5.00 per cent output growth was output per worker, and the wedge was the entire growth rate. In year 3, output per worker went from Rs 2,10,000/- to Rs 2,10,000/-, a rise of nil. The number at work rose 4.00 per cent. So the whole of the 4.00 per cent was people, and the wedge was zero. Check the identity both ways: 1.00 multiplied by 1.05 is 1.05, and 1.04 multiplied by 1.00 is 1.04. Both hold exactly.

A short warning about the shorthand saves an error later. People usually say output growth equals employment growth plus productivity growth, adding rather than multiplying. The addition is an approximation, and what it drops is the product of the two rates. In both Sankhya years that dropped product is exactly nil. One of the two rates is nil, and anything multiplied by nil is nil, so the addition is exact here. When both rates are large and positive the dropped product is not nil, and the addition understates. Addition serves when both rates are small and multiplication when they are not, and which of the two is in use should always be clear.

Which of the four possible combinations does Sankhya actually show?

Two counts, each able to rise or not rise, give four combinations rather than one. All four happen, all four are ordinary, and only the first of them is what most readers picture on being told an economy grew.

The first is output up and work up, the case everybody has in mind. More is produced and more people are producing it. The second is output up and work flat or falling, where output per worker carries the whole rise. The third is output flat or falling and work up. Output per worker is going backwards: more people are producing the same amount or less. People move into low output work when nothing else is available. The fourth is both giving way, where production and employment fall together.

Now put Sankhya into that grid, and the arithmetic does the arguing. Year 2 grew 5.00 per cent with the number at work stuck at 8.00 crore, and the unemployment rate rose from 4.76 to 6.98 per cent; year 3 grew more slowly at 4.00 per cent and the unemployment rate fell to 5.45 per cent, so the better year for output was the worse year for work. That is not a rhetorical flourish and it is not a special case constructed to be odd. The pattern is what happens whenever the labour force grows and the number of jobs does not.

Year 2 taken apart slowly contains the whole mechanism. Output rose 5.00 per cent, so more was genuinely produced. The number of people at work was 8.00 crore at the start and 8.00 crore at the end, so not one additional person found work. Meanwhile 0.20 crore more people started working or looking, some having turned old enough, some having finished studying, some simply deciding to start. The labour force went from 8.40 crore to 8.60 crore. Those 0.20 crore had to land somewhere, and the number of jobs did not move, so all of them landed in the unemployed count, taking it from 0.40 crore to 0.60 crore. The rate followed: 0.40 on 8.40 is 4.76 per cent, and 0.60 on 8.60 is 6.98 per cent. Every step of that is subtraction and division that can be redone on paper.

The three Sankhya years, both counts side by side

Every output, employment and labour force quantity in the table above belongs to the Republic of Sankhya. Each rate has been worked out again from the underlying counts rather than lifted from a summary, so any of them can be redone with a calculator to the identical figure. Participation is the labour force on a working age population of 13.20 crore.

Now the mirror image, and it is the half readers find harder to believe. Year 3 was the weaker year for output: 4.00 per cent against year 2's 5.00 per cent. Yet the number at work rose from 8.00 crore to 8.32 crore, an additional 0.32 crore people working. The labour force grew again, by the same 0.20 crore, from 8.60 crore to 8.80 crore. So 0.32 crore new jobs against 0.20 crore new entrants leaves 0.12 crore, and the unemployed count fell from 0.60 crore to 0.48 crore. The rate went from 6.98 per cent to 0.48 on 8.80, or 5.45 per cent. None of the growth that year was output per worker, so the same year brought slower growth, more hiring and a falling unemployment rate.

Four settings are written out in full below, so everything the panel teaches survives without a working browser. From year 1 at 5.00 and 5.00, employment growth is 0.00 per cent, the number at work stays 8.00 crore against a labour force of 8.60 crore, and the rate goes from 4.76 to 6.98 per cent, with output ending at Rs 16,80,000 crore and output per worker at Rs 2,10,000/-. From year 2 at 4.00 and 0.00, employment growth is 4.00 per cent, the number at work reaches 8.32 crore against a labour force of 8.80 crore, and the rate falls from 6.98 to 5.45 per cent, with output at Rs 17,47,200 crore and output per worker still Rs 2,10,000/-. From year 1 at 0.00 and minus 4.00, employment growth is 4.17 per cent, 8.33 crore are at work and the rate falls to 3.14 per cent: more people producing the same output. From year 1 at minus 2.00 and 0.00, employment growth is minus 2.00 per cent, 7.84 crore are at work and the rate rises to 8.84 per cent. The third setting is the one worth sitting with. The rate improves to 3.14 per cent while output does not grow at all, and that corner shows a falling unemployment rate is not by itself news of a stronger economy.

What does jobless growth mean, and does it settle anything?

Jobless growthA stretch in which output rises while the number of people at work does not, so the whole of the rise comes from each worker producing more. is the name given to the second corner of the grid: a stretch in which output rises while the number at work does not. Sankhya year 2 is exactly that. Output up 5.00 per cent, people at work unchanged at 8.00 crore, and the whole rise carried by output per worker moving from Rs 2,00,000/- to Rs 2,10,000/-.

The phrase describes an arithmetic pattern and nothing beyond it: output growth positive, employment growth at or below nil, and the wedge between them accounting for the whole difference. That is all it asserts. The phrase does not assert that the growth was fake, that the output was mismeasured, that anybody behaved badly, or that the year went badly.

Whether jobless growth is a problem depends on what happens next and on who is asking, and those are two different questions with two different kinds of answer. The next stretch is genuinely open: the same output produced by fewer people can mean more is left over to pay them, or it can mean the gains sit somewhere other than with the people who used to do that work, and the arithmetic determines neither. Who is asking matters just as much: a person already in one of those 8.00 crore jobs and a person among the 0.20 crore who joined the labour force that year experienced Sankhya year 2 as two completely different events, and neither of them is wrong about it.

Sharing out the gains is a question of its own, and the arithmetic does not answer it. The arithmetic settles whether the pattern is present, a prior step and a genuinely useful one. A reader who can look at a growth year and say the whole of it was output per worker has learned something checkable. A reader who then says the year was therefore bad has said something the numbers do not carry. The first claim holds; the second belongs elsewhere.

Which of the two counts matters more, and to whom?

The question gets asked constantly and has no single answer, and the absence of one is not a dodge. Different readers are asking different things and the same two series answer them differently.

A lender reads output first. Output is what generates the cash that services a debt. A lender assessing a borrower whose sales track the wider economy wants to know whether there will be more to sell into. The employment count matters to that lender too, but at one remove: it decides whether the borrower's customers can pay, and that shows up later in the same loan.

An analyst reads whichever of the two feeds the business in front of them, and it is usually neither of them cleanly. A company making construction materials lives off output. A company selling small daily items to wage earners lives off how many wage earners there are and what they take home, so the employment count is closer to the truth for it than the output figure is. An analyst who reads only the headline growth rate for both has used one number for two different businesses.

A household reads work, and nothing else comes close. A growth rate is an abstraction to a household; a job is money arriving on a specific day. In Sankhya year 2, a household with somebody among the 0.20 crore who entered the labour force and found nothing experienced a 5.00 per cent growth year as a bad year. The reading was correct: the growth reached them not at all.

An investor holding for years reads both, and reads the gap between them as information in its own right. A long stretch where output grows and employment does not shows the growth is coming from output per worker, which says something about where the rise is landing. A long stretch where employment grows and output barely does shows the opposite. Neither is a signal to do anything; both are descriptions of what kind of growth is happening.

A reader who wants one answer to which count matters more has asked a question that does not have one, and the useful move is to say which decision is being made before choosing a series. Name the decision and the right series usually names itself.

How should the two counts be read together?

Four habits, and they take about a minute once learned.

Never read one count off the other. A growth figure is not a hiring figure and a hiring figure is not a growth figure. A report that gives a rise in output and then states what it means for jobs without showing a jobs number has guessed, and the reason that guess is unreliable is now clear.

Look at what output per worker did. Output per worker is the wedge. If output grew 5.00 per cent and output per worker grew 5.00 per cent, employment growth was nil, and that follows without anyone having to report it. If output grew 4.00 per cent and output per worker was flat, employment grew 4.00 per cent. The third number is never independent of the other two.

Look at the labour force, not just the unemployment rate. A rate can improve because people found work or because people stopped looking. Sankhya year 2 shows the reverse case: the rate worsened while nobody lost a job, purely because 0.20 crore more people started looking.

Ask what a definition change would do before assuming a real change happened. The employment count has a fuzzy edge, so a shift in what counts as work can move the series without one person changing what they do. Where a series jumps at the exact point a survey changed, treat the jump as a question rather than an answer.

The Republic of Sankhya, the tailoring workshop, the farm and the tea stall are invented.
Educational material. Not advice on any investment, tax, budget or market position.