On Lost Decades

The missing variable in asset pricing

Economic theory derives the value of an asset based on its expected future earnings. While sensible, this only describes the situation partially: it tells you why one asset trades expensive relative to another, but says nothing about why everything is expensive or cheap at the same time.

All assets are held by people, directly or through intermediaries. It doesn’t matter whether an individual buys an ETF, invests through a fund, or hands money to a bank on terms. Corporates and banks are owned and funded by people too. And whether a person is a net buyer or net seller of assets at all isn’t a valuation decision. It’s a life-cycle decision. Workers with surplus income buy assets. Retirees with income deficits sell them. Valuation theory allocates the flow between assets. The direction and size of the flow itself is largely beyond that; more directly described by demographics than asset pricing theory.

A generation is largely synchronized in this. It accumulates together and it liquidates together, resulting in a net flow into or out of assets that valuation theory does not model. Saving and dissaving behavior of generations change slowly; a usable proxy is thus just the flow of people through the workforce: the growth rate of the working-age population. When more people leave the accumulation phase than enter it, the structural bid under all assets becomes a structural offer.

The data backs this up. Across 22 advanced economies from 1950-2010, property prices and broad-money demand tracked the working-age population as a common long-run driver¹. Across 22 OECD countries from 1970-2009, real house price growth ran opposite the old-age dependency ratio, at an elasticity of about -0.68, and with population size at about +1.05².

Three amplifiers and gates sit on top of the demographic base flow:

1. Liquidity of the absorbing generation. A young gen that lived through wage stagnation or a cost-of-living crisis can’t absorb the offered assets even at low prices.
2. Retirement reflexivity. The reversal feeds itself. Prices fall, and people counting on that money to retire stop feeling safe. So they sell now, before it falls further, and that selling pushes the price down more.
3. Storage medium. If the accumulating generation stored wealth in price-set assets (equities, land), the reversal discharges through prices. If it stored wealth in par-value claims (deposits, insurance, pay-as-you-go entitlements), the reversal hits via balance sheets and inflation instead; there is no price to crash.

Japan: the calibration case

Three generational facts set the stage:

• The dankai cohort, born 1947-49 at ~2.7M births a year (the 1960s managed ~1.6M), entered the workforce 1967-69 and hit peak accumulation age of around 40 in 1987-89. Which happen to be exactly the bubble years.
• Net workforce flow turned negative in the late 1980s; the pre-war birth cohorts leaving were larger than the 1960s birth-dearth cohorts entering. Working-age population peaked 1995, labor force 1998.
• The cohorts that were supposed to absorb all of this in the 1990s instead walked into the post-bubble hiring freeze, the “employment ice age”: a generation with abnormally little money facing an abnormally large offered supply.

So the structural bid under Japanese assets peaked and reversed at the turn of the decade. Price-to-rent hit 3-4x sustainable.

The needle was policy: The Bank of Japan raised the discount rate from 2.5% to 6.0% between May 1989 and August 1990, deliberately, to deflate asset prices. The Ministry of Finance capped real estate lending in April 1990. Equities peaked 29th Dec 1989. Land followed with a lag. The needle chose the date, but the reversal in flow chose the destiny. With the structural bid gone, there was no cohort on the other side to catch the falling market, and prices did not reclaim the peak until 2024. Thirty-four years.

Europe: past the threshold, crash pre-empted

Europe is not waiting for its Japan moment. It is inside it. Germany’s working-age population peaked in the late 1990s, and the EU aggregate crossed around 2007-2010. The symptoms are all there. The Euro Stoxx 50 needed 25 years to reclaim its 2000 peak. Italy has been a near-perfect Japan replay since 2007. German domestic assets went sideways for 15+ years, hidden in equities by the DAX being quoted as a total-return index (the price-only Kursindex tells the honest story).

Germany never had a visible burst, because Germans never kept their wealth in anything with a price attached to it in the first place.

Germany in detail. At its accumulation peak (~2000), German household financial wealth sat roughly:

Storage medium	Share	Price-set?
Bank deposits, Sparbuch	~35-40%	no
Life insurance, occupational pensions	~30-35%	no
Bonds, Bausparen	~10%	par at maturity
Equities and funds	~10-15%	yes
Pay-as-you-go pension claims	outside the balance sheet	not tradeable

That accumulation flow went almost entirely into nominal claims, not price-set assets. Banks and insurers exported the resulting savings glut, running a current account surplus of 5-8% of GDP for two decades and recycling it into foreign bonds and loans. The structural bid never inflated domestic asset prices, so the structural offer had nothing to deflate. Homeownership at ~47%, the lowest in the EU, kept even housing out of the storage role until the 2010s. Germany’s ~12-14% household equity participation rate is also among the lowest of any developed market, and it tracks almost exactly where German valuations cluster. Low participation, low multiples: more evidence that the flow into the market was structurally small, not just the flow out of it.

The liquidation leg discharges accordingly: deposit drawdown and insurance redemptions, met by running down bond books, with the ECB acting as guaranteed counterparty (2015-2022 QE absorbed more bonds than were net issued). The residual loss was delivered silently, through inflation. 2021-23 alone took roughly 15-20% in real terms off deposits and Lebensversicherungen. Same demographic hole Japan fell into. Germany just paid for it differently, through falling nominal claims and rising payroll contributions instead of a visible price collapse.

Japan lacked one further valve that Germany got to use: migration. The 2015 and 2022 waves effectively imported a missing cohort. The German housing boom of 2010-2022 lined up with them, helped along by zero rates.

China: should already have happened; partially did

On the model, China’s crash should already have happened at full Japan scale. Working-age population peaked somewhere around 2011-2014 and the labor force has been shrinking since. Households parked their wealth overwhelmingly in a single price-set asset: property, roughly 70% of household wealth. And courtesy of the one-child policy, the generation that has to absorb all of it is the smallest in this whole comparison, facing the largest offered supply. Every precondition is worse than Japan’s.

It partially did. Credit kept the bubble inflating for nine years past the crossover, until Beijing supplied its own needle in August 2020 with the Three Red Lines. Then the bill came due. Evergrande collapsed in 2021, prices have fallen 20-30% or more from their peak, consumer deflation set in, and youth unemployment hit record highs. But it’s all smaller than the model says it should be. On these fundamentals, a Japan-grade unwind implies far deeper marks than the official indices show.

China is cheating. The government leans on prices directly: whitelisting which projects get financing, buying up unsold units itself. Banks get told to extend and pretend instead of foreclosing. The losses don’t disappear, they just move onto the sovereign’s books instead of the market’s. It’s the same directed-credit playbook China runs everywhere else in its economy, the one that already flooded the world with EVs nobody asked for. Whoever’s politically inconvenient to let fail gets bailed out. Financial return has nothing to do with it.

Unclear which of two things this is. Maybe this is a clever move: instead of a price crash China gets a slow fiscal and inflationary trim; or maybe it’s just re-leveraging, delaying but not defusing a larger crash.

India: pre-crossover

India is on the other side of the curve entirely; the Japan-of-1965 position. Working-age population grows until roughly 2045-2050; median age 28. The multi-decade accumulation phase is just getting started. A growing worker cohort keeps bidding up domestic assets for a generation, and nothing like a crossover shows up before the 2050s. The real question is how much of that generation’s savings actually lands in stocks and real estate, versus gold and informal holdings, the way it traditionally has in India.

United States: pre-pop, and the valve is closing

The interesting case, and not a new one: Mankiw and Weil predicted back in 1989, from the same cohort-flow logic, that as the baby-bust generation reached house-buying years in the 1990s, US housing demand growth would slow to its lowest rate in forty years³. The US native-born working-age population stopped growing around 2012; every year since, the crossover has been deferred by exactly one thing: immigration.

That stream has now been cut. 2025 saw net negative immigration and 2026 projected to stay in negative territory⁵. If that holds, the US crossover is happening right about now.

And the US has none of Germany’s insulation. Americans keep their wealth in things with prices on them: roughly 60% of households own stocks, retirement means a 401k marked to market rather than a pay-as-you-go claim, and the house is the second pillar. When the reversal comes, it comes fully through prices of those.

Apply the Japan/China mapping (crash within ±5 years of crossover) and you get a dangerous window at ~2027-2032.

Two caveats to this thesis:

1. Capital immigration. US assets are uniquely absorbed by foreign savers; the rest of the world holds on the order of $30T+ of US securities, and the accumulating cohorts of other countries, including Europe’s young ETF generation whose savings plans are ~70% US, function as an external absorbing generation no other market in history has had. The demographic crossover is domestic; the bid is global. For the model to produce a crash, the foreign leg has to fail simultaneously, via dollar aversion, geopolitical decoupling, or the foreign cohorts hitting their own liquidation phases. Timing is also relevant: Europe and China have their own retirement waves coming, so the external bid weakens through the 2030s for reasons that have nothing to do with the US.
2. Negative net supply and mechanical inflow. US corporates retire roughly $1T/yr of equity through buybacks, shrinking supply just as households are selling. And 401ks keep buying regardless of price. Both mute the reversal without eliminating it.

Verdict. The spring is loaded. If migration stays shut, the crossover is underway, the wealth sits in price-set assets, retirement reflexivity is maxed out, and there’s no domestic par-value channel to bleed the pressure off through.

Whether it releases as a crash or bleeds out as stagnation depends on whether the pin lands while both valves, migration and foreign absorption, are closed at once. That conjunction is plausible, but it isn’t the base case within the window. The base case is that foreign and mechanical bids convert the reversal into structural underperformance, with a Japan-grade event as the tail outcome if a credit or rate pin arrives during a period of dollar aversion. Reopen immigration after 2028 and the spring gets defused within years. It’s the only crossover in the dataset that’s a policy variable rather than a destiny.