On Tuesday a shopping mall collapsed in southern Japan — a mall that had reopened just last month, after ten years of rebuilding from the last big earthquake in the same region. Thirty-four people are dead or feared dead across Kumamoto, a number still being confirmed.

Japan rehearses this day better than any country on Earth. And yet, across Japan, roughly two homes out of three have no earthquake insurance at all. Not carelessness, not poverty — a design, written in 1966, that built the world's most elaborate earthquake-insurance machine and decided, in the same pages, exactly how incomplete it would be.

Tuesday, 28 July, late afternoon: a magnitude 7.1 quake strikes Kumamoto Prefecture — shallow, the dangerous kind. The mall in Kashima partially collapsed; ~3,000 shoppers were evacuated before a gas explosion brought down more of the structure. Eight workers died at a paper mill; bullet trains stopped; the castle walls — still under repair since the last disaster — fell again. Ten years ago the same prefecture was hit by a quake that killed nearly 300 people counting the aftermath; the region was still finishing that recovery when Tuesday arrived.

For days the cameras will count what earthquakes make us count: the dead, the missing, the 72-hour window in which survivors can still be found. But a second disaster arrives weeks later, in envelopes — repair estimates, demolition orders, bills. Who pays them was decided sixty years ago.

In 1964 an earthquake flattened parts of Niigata, and Japan's insurers faced honest arithmetic: one big quake could produce claims larger than the industry's entire assets. For two more years, home earthquake insurance simply did not exist. Then the 1966 Act built something remarkable: every household earthquake policy, from any insurer, is passed on — 100% — to a single company (JER), which passes most of it upstream to the government. In the worst possible disaster the split is roughly: insurers 3%, JER 10%, the state 87%. Nothing is reinsured abroad: Japan holds its own household earthquake risk entirely at home.

The law states its purpose in Article 1: to spread earthquake insurance and thereby contribute to "the stability of the lives of disaster victims." It has paid after every major quake for sixty years and never failed. Which is what makes the next two design choices so interesting — because the same law decided, precisely, how much protection it would refuse to sell you.

What the law lets you buy: earthquake cover exists only as an add-on to a fire policy — never on its own. The insured amount is set by statute at 30–50% of the fire policy's value, with hard ceilings (about ¥50 million for a building, roughly $300,000), and payouts come in fixed steps by damage category, not as your repair bill. Deliberate — the product exists "to restart a life, not to rebuild a house," because promising to rebuild every house is what bankrupts a system in a megaquake.

Follow the incentive downstream: full premiums for half a house, paid in steps — many households conclude it isn't worth it. Of those who buy fire insurance, about seven in ten add the earthquake rider; across all of Japan's homes, only about 35% carry earthquake cover. An under-insured product, truncated by law to keep the system alive, produces an under-insured nation.

The whole system carries a legal maximum per earthquake: ¥12 trillion (~$80 billion), a number parliament reviews and raises; today it is calibrated to survive a repeat of the 1923 Great Kanto earthquake. If claims ever exceed it, the law is explicit: every payout may be reduced pro-rata until the total fits. The system cannot go bankrupt — because your check is allowed to shrink. The government pledges it would act beyond the framework in such a disaster; that pledge is effort — the reduction clause is law.

And the quieter force: after every disaster the state helps everyone — grants, rebuilt infrastructure, support flowing to insured and uninsured alike. A family deciding whether to buy the rider sees the uninsured neighbour's help arrive anyway; the voluntary product competes with a free, implicit one — and loses. So coverage climbs only in steps of fear: 10% before Kobe 1995, 35% today.

Attach by default (New Zealand). Disaster cover fuses automatically to every home fire policy, funded by a small charge — nobody opts in, so nobody is out. The cost came due in Christchurch: a city of auto-covered homes produced a flood of claims assessed house by house; settlements dragged for years. Default solves who pays — not how fast.

Make it mandatory (Turkey). After Izmit 1999, earthquake insurance became legally compulsory. The cost is the gap between a law and a habit: checked mainly at property sales, roughly half of homes comply, fewer in the poorest regions; in 2023 the payouts were real but small. A mandate without enforcement is a volunteer program with extra paperwork.

Leave it to choice (California). A genuine choice — no default, no mandate, no deep guarantee: about one home in ten is covered. Remove Japan's cathedral and you get a third of Japan's coverage. The truncation is why coverage stops at 35%; the cathedral is why it climbs that high at all.

Pay fast, not exact (parametric). A fixed sum within days, triggered by measured shaking — no assessors, no envelopes. The cost is basis risk: the trigger measures the ground, not your house. Speed bought with accuracy — a complement, not a replacement.

Back to the mall — rebuilt across ten years, reopened for one month, shaken again on a Tuesday. Japan will rebuild it; the engineering question was answered decades ago. The financial question was answered too, more quietly: a machine built to survive anything — and the same law capped the product at half a house, priced two-thirds of families out of wanting it, and reserved the right to shrink every check in the one megaquake the machine exists for. There is no villain: in 1966 this was the only honest deal available, its limits written in plain text.

But it leaves a question for every country's safety systems — flood, pension, deposit insurance, all of them. When the state guarantees a system will survive the disaster, ask the follow-up: what exactly is guaranteed to survive — the system, or your check?