The Shape of People to Come

0 · The fixity principle

Every person who will be 30 years old in 2045 has already been born. There is no policy, no boom, no crisis that changes this. A government can print money; it cannot print 25-year-olds. Once a generation is on the books, only two forces ever edit it: migration moves people between ledgers, and mortality retires entries early. Everything else — every workforce, every tax base, every consumer market of the 2040s — is already standing in today's kindergartens and cubicles, countable to within a rounding error.

This is what makes demographics unlike every other macro variable. GDP forecasts five years out are famously embarrassing. Population-structure forecasts twenty years out are boring — in the best possible sense. The UN has published population projections since the 1950s, and for the already-born cohorts their medium-variant errors are small single digits. When someone tells you what China's labor force will look like in 2040, they are not speculating; they are doing arithmetic on people who already exist.

1 · Pyramid, column, urn

Plot a country's population by age — youngest at the bottom, men left, women right — and you get its demographic silhouette. For nearly all of human history, every country had the same one: a pyramid. Lots of children at the base, attrition all the way up. High fertility, high mortality; a society perpetually young, perpetually replacing itself.

Industrialization breaks the pyramid, and it breaks it in a specific order. First mortality falls (sanitation, antibiotics, food security) while fertility stays high — the base bulges, population explodes. Then fertility falls as people urbanize, and the silhouette straightens into a column: cohorts of roughly equal size, a population that neither grows nor shrinks much. Hold fertility below replacement for another generation or two and the base narrows under the bulge: the column becomes an urn — an inverted pyramid balancing on a thinning stem of young people, top-heavy with retirees.

The three silhouettes are not three destinies; they are three timestamps on the same road. Nigeria is where South Korea was in 1965. Japan is where South Korea will be in 2045, except South Korea is traveling faster than any country ever has. The only question a demographic profile really asks is: where on the road are you, and how fast are you moving?

2 · The order that industrialized everyone

Why did the whole world start down that road, and why do countries sit at such different mileposts today? The mechanism is urbanization, and the timetable was set — mostly by accident — at a New Hampshire ski resort in 1944.

On a farm, a child is free labor with a 6-year lead time: another set of hands for weeding, milking, hauling water. In a city apartment, that same child is a 25-year unfunded liability — an extra bedroom, school fees, university, a delayed career for a parent. Nothing dissolves fertility like moving the family from the first environment to the second, and industrialization is precisely that move at civilizational scale. Every country that industrializes urbanizes; every country that urbanizes watches its birth rate fall. No exceptions have been observed. The correlation survives religion, regime type, and family policy; governments have thrown every pro-natal incentive ever invented at it and bent the curve by decimal points.

The Bretton Woods settlement is what made the move near-universal. The American offer to its Cold War alliance — we patrol the oceans, our market is open, sail anything anywhere and it arrives — decoupled industrialization from empire. You no longer needed a navy and colonies to get inputs and reach customers; you needed a port. So industrialization, which took Britain a century and a half, was compressed into ~70 years for Germany and Japan's reconstruction, ~40 for Korea and Taiwan, and roughly 30 for coastal China. Each compression compressed the fertility collapse too. The countries that boarded the escalator late are riding it fastest — which is why South Korea went from TFR 6 to TFR 0.75 within a single working lifetime, a descent no society has ever experienced before.

3 · The price of a person: capital structure by age

To convert silhouettes into economics you need one idea, and it is fifty years old: people run a lifelong mismatch between what they produce and what they consume. Modigliani called it the life-cycle hypothesis — we borrow young, save in midlife, dissave old. The National Transfer Accounts project later measured it, age by age, in dozens of countries. Define labor income \(y_l(a)\) and consumption \(c(a)\) at age \(a\); the gap

\[ \mathrm{LCD}(a) \;=\; c(a) - y_l(a) \]

is the life-cycle deficit. It is positive twice (childhood, old age) and negative once (working life). Somebody must finance both deficits out of the one surplus — through family transfers, through the state, or through assets. A country's macroeconomy is, to first order, the sum of its citizens' positions along this curve.

Walking the curve, tranche by tranche:

Two rows deserve a second look. The financiers (46–64) are the pivot of the whole system: they simultaneously supply the tax base and the investment capital. And the drawers (65+) don't just stop contributing capital — they invert their portfolio. A 55-year-old buys equities and corporate bonds; a 75-year-old holds government paper and cash, and sells assets to eat. Aging doesn't merely shrink the pool of capital; it changes what the remaining capital is willing to finance. Risk capital — the kind that funds startups, factories, and emerging markets — is disproportionately a product of societies with fat 45–64 cohorts.

4 · What the shape does to an economy

Sum the life-cycle curve over a population and the silhouette starts writing macroeconomic history. This article ships two support ratios, because the honest one and the intuitive one are not the same number. The intuitive one is the headcount old-age support ratio — workers per retiree, ignoring children:

\[ SR_{\text{head}} \;=\; \frac{\text{population aged } 20\text{–}64}{\text{population aged } 65+} \]

The textbook-correct one is the economic support ratio of the National Transfer Accounts literature: weight every person by how much a person of that age actually produces and consumes — the very curves of Figure 3 — and divide effective producers by effective consumers of all ages:

\[ SR_{\text{econ}} \;=\; \frac{\sum_a y_l(a)\,P(a)}{\sum_a c(a)\,P(a)} \]

The two disagree in an instructive way: Nigeria has the best headcount ratio on the map (≈15 workers per retiree) and nearly the worst economic one (≈0.57 effective producers per effective consumer) — because children consume without producing, and the headcount ratio simply doesn't look at them. One methodological honesty note: measured NTA age profiles exist for only some countries, so — following Mason's (2005) original global implementation, which used a single Taiwan-1998 profile for every country — this site applies one stylized global profile pair (exactly the Figure 3 curves) everywhere. The atlas discloses this wherever \(SR_{\text{econ}}\) appears.

Why the fuss over a denominator? Because the support ratio is where growth accounting meets the pyramid. Income per effective consumer factors, as an identity, into productivity times the support ratio, so its growth decomposes as

\[ g\!\left(\tfrac{Y}{N}\right) \;=\; g\!\left(\tfrac{Y}{L}\right) \;+\; \tfrac{d}{dt}\ln SR_{\text{econ}} \]

The second term is the first demographic dividend (UN Population Division Technical Paper 2017/1): it exists exactly while the support ratio is rising, adds to growth one-for-one while it lasts, and is an accounting identity — not a causal promise. When fertility first falls, a country gets its golden decades: the huge final pre-collapse cohorts flow into working age while the number of children they must support shrinks and the retirees above them are still few. Bloom and Williamson credit this phase with as much as a third of East Asia's growth miracle; Lee and Mason estimate the favorable window typically runs five decades or more (Europe ~1962–2000, Asia ~1975–2033) before the term flips sign. Then those same workers, having few children to spend on, save — and their savings become the deepest pool of investment capital the country has ever seen: the second dividend. Cheap labor and cheap capital at once. Growth looks unstoppable. Books get written about the national model.

The bill arrives on a schedule you can read off the pyramid. The dividend cohorts retire; the support ratio dives; the savers become sellers. Domestic consumption sags (retirees buy services and medicine, not apartments and sedans), so industrial capacity built for a bigger, younger home market must export or die — which is why the world's aged industrial economies (Germany, Japan, China, Korea) all run or ran structural surpluses aimed at whoever still has consumers. Their surpluses are not a strategy; they are a symptom. Meanwhile the fiscal engine reverses: the financier tranche that funded the state shrinks as the drawer tranche that bills it swells.

One timing correction the popular tellings get wrong: the consumption drought is not about losing twenty-somethings. Look back at Figure 3 — consumption peaks in mid-life, because spending follows income and income peaks after 45. A shrinking birth cohort therefore takes decades to dent aggregate demand; the drought arrives when the prime-age tranche thins, one generation after the birth dearth. (Zeihan's version puts peak consumption in the 20s–30s, which moves his timelines earlier than the age profiles support — see §9.)

The trade channel isn't a metaphor; it's measured. Higgins (1998, International Economic Review) showed that age structure moves savings and investment differently — investment demand tracks the youth share, savings supply tracks the mature-adult share — so a country's position in the transition predicts whether it imports or exports capital, with demographic effects on the current account exceeding six percent of GDP over three decades for a number of countries. The IMF institutionalized the channel: its External Balance Assessment models the current account with a demographic block (population growth, old-age dependency, prime-aged-saver share, plus a longevity term interacted with dependency twenty years ahead) that has survived every robustness pruning since — the official arithmetic behind "aging exporters must run surpluses." Both specifications use exactly the age shares this atlas computes.

5 · Beyond birth rates: a better instrument panel

Public discussion of all this runs on one number: the total fertility rate. TFR is a fine thermometer and a terrible dashboard. It tells you about one flow at one edge of the pyramid, thirty to sixty years before that flow matters economically — and it says nothing about the stock: the workers, taxpayers, savers and retirees who are already here and already scheduled. Korea's crisis-grade TFR of 0.75 will not touch its labor market until the 2040s; the workforce contraction Korea faces this decade was locked in by the birth rates of the 1990s. If you want to read a country's economic future off its population, you want instruments that look at the whole silhouette. This article computes six, for every country, from the single-year age data:

No single one of these is novel; the point is the panel. A country can look fine on TFR and terrifying on WRR (immigration-fed economies with aging natives), terrifying on TFR and fine for another decade on SR (China until ~2032), or best- in-class on the headcount ratio and nearly worst on the economic one (Nigeria, whose children the headcount simply doesn't count). The instruments disagree in informative ways, and the map in section 7 lets you watch them disagree.

6 · A composite demographic-strategic score

Demographics is the engine, but engines run on terrain. A young population inside indefensible borders (the Sahel), or an old population sitting on the world's best farmland behind two oceans (the US Midwest, give or take the "old"), are different propositions. Following the geostrategic tradition — Zeihan is the loudest recent voice, but the furniture goes back to Mackinder and Spykman — the atlas layers five contextual components onto the two demographic ones:

1. Demographic structure (computed): the average of the standardized support-ratio, median-age and POADR signals.
2. Demographic trajectory (computed): standardized WRR, built-in growth, TFR and capital-shift — where the structure is heading.
3. Geographic security (authored ✎): moats, plains, chokepoints, strategic depth.
4. Agricultural potential (authored ✎): can it feed itself if trade gets expensive?
5. Energy & resources (authored ✎): domestic energy vs. import dependence.
6. Neighborhood & trade resilience (authored ✎): who are the neighbors; what happens to your model if long-haul shipping stops being free?
7. Capital access (authored ✎): market depth, reserve status, sovereign balance sheet.

The computed components standardize each raw metric across all 236 countries (winsorized at the 5th/95th percentiles, signed so that younger/stronger is positive, clamped to ±3):

\[ z_{c,m} \;=\; \mathrm{sgn}_m \cdot \frac{x_{c,m} - \mu_m}{\sigma_m}, \qquad S_c \;=\; \frac{\sum_{j \in A_c} w_j \, z_{c,j}}{\sum_{j \in A_c} w_j} \]

where the weights \(w_j\) live on a simplex (sliders renormalize to sum to 1) and \(A_c\) is the set of components a country actually has data for — missing components are excluded and disclosed, never imputed to zero. Authored rubric scores (0–10) map linearly onto the same scale via \(z = (s-5)/2\).

One measurement caveat baked into the demographic components: two of the three structure signals hard-code age 65, and remaining-years-based measures consistently show a slower pace of aging than fixed-age measures — that is Sanderson & Scherbov's central finding, and it is why the POADR (which lets "old age" start at 74.5 in Japan and 60.5 in Nigeria) sits in the panel as the corrective. The finding is conditional, not universal: it holds where life expectancy is rising (nearly everywhere under the UN medium variant) and fails where it stagnates, as in 1990s–2000s Russia. The gap is large and widening in exactly the rich countries this article foregrounds — OECD-average remaining life expectancy at 65 is already ~18.5 years (men) / ~21.6 (women), 3.5 to 6.6 years past the 15-year threshold, and projected to rise roughly four more years by 2065.

7 · The atlas

8 · Ten portraits

Ten countries that between them cover the whole road from pyramid to urn. Each link opens the full workup in the atlas above.

9 · What this lens misses

A forecast tool is only as honest as its list of failure modes. Start with the framework this article set out to test-drive. Here is the Zeihan causal chain, link by link, graded against the primary evidence — SUPPORTED where it's textbook demography, CONTESTED where serious sources disagree, WRONG where the data says otherwise:

And here is where demographic determinism — mine included — breaks:

• Aging hasn't lowered growth — yet. Acemoglu & Restrepo (2017): across countries since 1990, "there is no such negative relationship in the data — if anything, countries experiencing more rapid aging have grown more." Their explanation: scarcer middle-aged labor accelerates automation, and demographics alone explain nearly half the cross-country variation in robot adoption. Every scary ratio above has workers in the denominator, and the denominator is exactly what the aged, capital-rich economies are engineering around (Japan and Germany lead the US in robot density for a reason).
• The asset-meltdown premise is weak in data. Poterba (2001, 2004): no robust link between age structure and returns on stocks, bonds, or bills in US/UK/Canadian time series; age-wealth profiles rise steeply in the 30s and 40s but decline only gradually in retirement; projected US asset demand doesn't fall through 2050. The §3 toy model's capital-scarce leg is theory ahead of evidence — "modest support at best."
• Migration is policy, and policy moves faster than birth rates. The age structure of the already-born is fixed; which ledger they're on is not. The US, Canada and Australia have effectively outsourced their fertility problem to the admissions queue. A country that can attract and integrate 25-year-olds can repair a thin cohort in years, not generations — and one political cycle can shut the valve.
• Retirement is a social convention wearing a biological costume — and it's already moving. The 65+ threshold in the headcount ratio is an artifact of Bismarck's actuaries; health at a given age keeps improving (that's why POADR is on the panel). The adaptation is legislated, not hypothetical: the OECD average normal retirement age rises to 66.4 (men) for workers entering in 2024, with 70+ already on the books in Denmark, Estonia, Italy, the Netherlands and Sweden.
• The institutions that price this professionally price strain, not collapse. The US intelligence community's Global Trends 2040 files demographics under chronic fiscal and governance pressure — not deglobalization or trade collapse. The OECD projects public pension spending rising from 8.8% to 10.0% of GDP by 2050: a real bill, roughly one Italy-sized policy problem per generation, not an apocalypse. When your thesis is more catastrophic than the reinsurers', the burden of proof is yours.
• The UN's track record cuts both ways. World-total projections have been remarkably good — most vintages within 1–2% even decades out. But the age-structure components this article runs on have historically carried larger errors (Europe's children overestimated, its elderly undercounted), and end-of-century figures swing half a billion between revisions. Trust the 2050 working-age counts (already born); discount the 2080 tails.
• The authored layers can be wrong. The geography scores encode a particular strategic worldview — that the maritime order frays and self-reliance gets repriced. If the next thirty years look like the last thirty, the "Zeihan lens" systematically overrates fortress economies and underrates trading hubs. That is exactly why the pure-demographics preset exists.
• Countries are not the only unit. Sub-national structure matters (rural Japan vs. Tokyo are different planets), and this version aggregates it away. A US-state-level layer is the natural next build on this data schema.