India's Demographic Dividend: The Opportunity and the Condition

A demographic dividend is a stretch of years in which an unusually large share of a country's people sit in the working ages relative to the people they support. A demographic dividend is an opportunity and not an outcome: it arrives only where that generation finds work productive enough to carry the dependents on both sides of it, and the stretch of years ends on a schedule nobody chooses.

Two words, and almost all the argument sits in the second one. A dividend is something paid out. Nothing about an age structure pays anything out on its own, any more than a shop full of stock is a shop full of takings. The stock has to be sold. The generation has to be put to work. The part of the phrase that is settled and the part that is not are two different things, and most of the confusion about a demographic dividend comes from running them together. The settled part is the one everybody quotes.

What is actually happening to an age structure when people call it a dividend?

The arithmetic is identical in a household and already familiar there, so a household makes it plain. A household of six with one earner is stretched. The same household ten years later, when three of the children have finished studying and taken work, has four earners and two dependents. Nothing about the household got richer by ageing. The ratio of the people bringing money in to the people the money has to cover changed, and the ratio changed because of who had a birthday, not because of anything anyone decided.

An economy does the same thing on a larger scale, and slowly. The age structureHow a population is spread across ages, counted in bands such as under fifteen, fifteen to fifty-nine and sixty and over, rather than one year at a time. of a country is the count of its people arranged into age bands. Every country supports people outside the working ages: children who are studying, the retired, and those who cannot work. When a large cohortEverybody born within the same stretch of years, followed as one group as it moves up through the ages. A cohort keeps its size, minus deaths and migration, for its whole life. reaches the working ages and the cohort born after it is smaller, two things happen at once: the working age populationThe count of people who fall inside the ages a country treats as working ages. The count uses age alone, and says nothing about whether any of them has work. swells, and the number of children behind it stops growing. The share of everyone who sits inside the working ages rises. The stretch of years over which it keeps rising is what gets called the windowA stretch of years with a start and an end, after which the condition that opened it no longer holds. Nothing about a window guarantees that anything was done during it..

All of that is arithmetic about ages, and nothing else at all. No output has been produced, no work has been found, no income has been earned. A birth in one year and a sixtieth birthday in another have been counted, and a proportion has been taken. The claim at this stage is thin. Every interesting thing about a demographic dividend happens after this point rather than at it.

Why is a demographic dividend an opportunity rather than an outcome?

Because a share of a population is a supply of people, and a supply of anything only becomes something when it is used. A larger working age share raises the output of an economy only if those extra people are actually in work. And a larger working age share raises output per personTotal output divided by every person in the country, dependants included. Output per worker is a different number, dividing by the people in work only., which is the number a dividend is really a claim about, only if the work those people find produces enough to be worth doing.

Two things therefore have to be true, and neither of them is settled by anybody's age. Whether people of working age are in work at all is a question about the labour force and how much of it is employed, treated separately under employment. Whether the work they find produces much is a question about output per worker, treated separately as well. A demographic dividend is a term inside a growth calculation rather than a growth rate, and treating it as a forecast is the commonest error made about it. The age structure hands an economy a bigger number of possible workers. A bigger number of possible workers is the whole of what it hands over.

Think of a wedding hall that has just doubled its bookings capacity. The capacity is real, it is measurable, and it is on the premises. Whether takings double depends on whether the dates fill and on what each booking pays. Nobody would look at the new floor plan and write next year's revenue from it, and yet that is exactly the move made when a projected working age population is turned into a projected growth rate.

What is the condition that the phrase usually drops?

Say the whole thing out loud and the missing half becomes obvious. A demographic dividend is the gain available if an unusually large working age generation finds work productive enough to carry the dependents on both sides of it. The gain is the part that gets quoted. The condition is the part that gets dropped, and dropping it turns a conditional statement into a promise.

The condition has three pieces, and all three have to hold together. The generation has to find work at all, and that requires somebody hiring at the scale the generation arrives in. A dividend is a claim about output per person, and a worker who barely covers herself leaves nothing over for the child and the retired parent counted beside her. So the work has to produce enough per worker to cover more than the worker. Output per worker is not a quality of a person but of a person working with something. For the work to produce that much, then, the tools, the premises, the electricity, the roads and the skills have to already exist.

A large working age population that does not find productive work is not a smaller dividend, it is not a dividend at all, and it is a harder problem than a smaller working age population would have been. The usual framing is inverted. The same arithmetic that promised more supporters per dependent instead delivers more people needing support, from an identical age structure, with no error anywhere in the counting.

Why does the window close, and what sits on the far side of it?

The window closes because the generation that opened it does not stay in the working ages. It ages out. Everybody in a large cohort has a sixtieth birthday roughly forty years after they had a twentieth one, and when they do, the band they were swelling empties into the band above it. The share that had been rising starts falling. Nothing intervened and nothing failed; the same process that opened the window closes it, and it was always going to.

Now watch where the window closes to. Almost nobody carries that part away. The dependency ratioThe count of people outside the working ages set against the count inside them, usually written as so many dependants per hundred people of working age. that improved does not return to where it started. The ratio goes past it. The large generation is now retired and being supported, and the generation doing the supporting is the small one born during the window, kept small by the fertilityThe average number of children born per woman in a population. Fertility is what decides how large the next generation will be, and it feeds through to the age structure decades later. that fell as the window opened. A big group being supported by a small one is a worse arrangement than the moderate group supported by a moderate one that existed before any of this began.

The closing of the window is not a risk to be managed but a certainty to be planned around, and the only variable in the whole sequence is what was built while it was open. On the invented Sankhya path drawn below, the working age share opens at 66.00 per cent, peaks at 70.00 per cent, and settles at 64.00 per cent once the large cohort has retired. Two points of working age share above where it began, then two points below. The peak is temporary and the level afterwards is not.

Which part of the growth arithmetic does a dividend actually move?

Total output can be written two ways at once: as a single figure, or as a count of the people in work set beside the average each one of them turns out. Splitting any total into a count and an average and then multiplying gives back the total it began with, so the second version claims nothing. The second version buys something else. A rise can now be shared out between the two halves, and those shares say something a single growth rate hides completely.

Run that decompositionSplitting one number into the separate parts that add back to it. The split shows which part moved. The parts must reconcile exactly to the total, or the split is wrong. on the invented Sankhya window and the result is stark. Output opens at Rs 17,47,200 crore and closes at Rs 27,16,560 crore, so the distance covered is Rs 9,69,360 crore. Of that, Rs 1,93,200 crore comes from a higher share of the existing working age population being in work, Rs 3,23,400 crore comes from the working age population itself being larger, Rs 3,49,440 crore comes from each worker producing more, and Rs 1,03,320 crore is what the first two changes produce jointly that neither produces alone. The four parts add to Rs 9,69,360 crore exactly.

The demography supplies one of those four bars and only one: a larger working age population raises the count of possible workers and adds nothing whatsoever to what any of them produces. An economy leaning on that bar is growing on more hands rather than better output per hand. A count of people can double, and then it has doubled and cannot do it again from the same source. A demographic dividend is therefore a temporary source of growth by construction, and that is its definition rather than a criticism of it.

Can one age structure produce three different outcomes?

The whole argument, worked on one invented economy. The Republic of Sankhya, an invented country, has about 20 crore people when the window opens. Working age covers 13.20 crore of them, or 66.00 per cent of everyone. Of those, 8.32 crore are in work, and each produces Rs 2,10,000/- of output in a year, so total output is Rs 17,47,200 crore and output per person, across every one of the 20 crore, is Rs 87,360/-.

Now run the window. In all three cases below the demography is identical and fixed: the working age population reaches 15.40 crore out of a total of 22.00 crore, so the working age share reaches 70.00 per cent, four points above where it opened. Nothing in any scenario touches that. Only the condition varies.

Read the last column and then the last row. The demography was identical in all three lines and the outcomes are not comparable to one another. Scenario one lifts output per person by 41.35 per cent. In scenario two every extra person found work but the work produced no more per head. Total output lifts by 29.57 per cent and output per person by only 17.79 per cent. The extra output had to be spread across two crore more people as well. Scenario three lifts total output by nothing at all and output per person falls 9.13 per cent, on a working age share that rose four full points.

Scenario three is the line to hold on to. The fall is not a failure of measurement or a rounding artefact. The working age share genuinely rose, exactly as any projection would have said it would, and output per person genuinely fell. India's own population and workforce figures are published by the bodies named below.

If the age structure is already known, what should be watched instead?

Ages move at one year a year and cannot be hurried. Everybody who will be of working age twenty years from now has already been born. The demographic part of this question is therefore the most predictable thing in the whole of economics. A projection of a working age population two decades out is not a brave forecast but something close to bookkeeping, and it surprises nobody who has looked.

The demography is the least uncertain part of the question and is therefore the least informative part of it. Information lives in what could turn out differently. Two things here could turn out differently. The share of working age people in work may rise or may not, and output per worker may rise or may not, and those two terms are exactly the ones nobody can project with any confidence. A reader who spends their attention on the age structure has spent it on the settled half and left the unsettled half unexamined.

Two series are worth watching, with the third treated as background. The first is whether the count of people in work is rising as fast as the working age population is. If it is not, the gap is accumulating as people of working age without work. The second is whether output per worker is rising. A rising count of workers with flat output per worker produces scenario two above, where total output climbs respectably and output per person climbs far less. Neither of those two questions is answered by an age structure: employment and how a labour force is counted, and output per worker in its own right, are each treated separately.

What does a large working age population mean for a business?

A lender or an equity analyst looking at a country with a swelling working age population is looking at two things at once, and they pull in opposite directions. A manufacturer sees a labour supply: more people available, wage pressure that stays contained, and the possibility of building capacity that would be unaffordable elsewhere. A consumer business sees a market: more households forming, more first purchases, more of the spending that comes with earning.

A market is made of income and income is made of productive work, so which of the two a large working age population turns out to be depends entirely on the condition. The manufacturer's case survives scenario three: labour is available and cheap precisely because so much of it is idle. Output per person fell in scenario three, so the consumer business's case does not survive it at all. A population that is larger and poorer per head is not the market anybody underwrote. An analyst who writes the same demographic paragraph into both cases has written it correctly for one of them and wrongly for the other.

The practical version, for anybody sizing a market from a population: get the demographic line out of the model as an input and put it in as a constraint. The demographic line sets the maximum. The share in work and what the work produces fill the space between the maximum and the actual, and those two get their own ranges, their own downside case and their own explicit assumption. A household deciding whether the local shop can support a second branch does the same thing without the vocabulary: it asks not how many people live nearby, but how many of them have money coming in.

Who publishes the population and work figures left out here?

Three bodies issue the material a reader will want next. The national accounts of India, in which output for the country is compiled, are issued by the Ministry of Statistics and Programme Implementation. The survey work from which India's employment and workforce estimates are drawn is carried out by the National Statistical Office. And the Economic Survey, in which structural questions of the kind treated here are set out and argued at length, is issued by the Ministry of Finance.

One caution applies before any of that is looked up. Ask each body what its series covers and which reference period it belongs to, and read that coverage note as carefully as the number itself. Population counts and workforce estimates come out of different exercises, with different coverage and different reference periods, and in some cases a different issuing office again. Establishing which body issued the particular series in hand is not pedantry: getting it wrong is how a reader ends up setting two series against each other that were never built to be compared, and then reading a difference between them as a finding. A number lifted out of a compiled series quietly stops matching its own source the first time that source is revised, and revising a compiled series is ordinary practice rather than a sign that anything went wrong.