The Great Unbunkering

Sanyo Securities filed for protection under the Corporate Reorganization Act on 3 November 1997, the first listed Japanese securities firm to do so, with liabilities reported at roughly 374 billion yen (about $2.3 billion). The next day an asset-preservation order produced the first default in the postwar history of Japan's interbank call market. The liabilities were not the point. The point was that an overnight loan between two Japanese financial institutions could simply fail to come back.

Two weeks later, on 17 November, Hokkaido Takushoku Bank abandoned self-rehabilitation and announced the transfer of its business. The Bank of Japan's Policy Board decided the same day on special lending under Article 25 of the Bank of Japan Act, in an amount described only as the minimum necessary to keep honoring withdrawals. Takushoku, founded in 1900, remains as of 2026 the only city bank to have failed in the postwar era.

On 24 November 1997 an extraordinary board meeting at Yamaichi Securities resolved to suspend operations toward voluntary closure and dissolution. The Bank of Japan's Policy Board decided that same day to extend a special loan under Article 25, channeled through Fuji Bank, to fund an orderly wind-down; the amount was stated only as the minimum necessary, and the decision was not made public until 12 December. Yamaichi had been founded in 1897 and was one of the four large securities houses. It failed in its hundredth year.

The press conference at the Tokyo Stock Exchange began at 11:30 that morning and ran more than two hours. The president, three months into the job, broke down and said the fault lay with management and that the employees had done nothing wrong. Liabilities were reported at roughly 3 trillion yen (about $18 billion) and about 7,500 people lost their jobs. The internal investigation published in April 1998 put the off-balance-sheet liabilities built up through tobashi — losses parked in paper companies — at about 264.8 billion yen ($1.6 billion), some 158.3 billion ($966 million) of it domestic and 106.5 billion ($650 million) overseas, a figure most reporting at the time rounded to about 260 billion.

The government wrote the lesson down in July 1999. That year's Economic White Paper, describing the period from spring to autumn 1998, acknowledged that as the economy contracted and banks tightened credit, companies conducting normal business operations — not only those carrying bubble-era assets — ran short of funds, and that some were driven into bankruptcy or closure. The white paper is careful. It says there were such cases. It does not count them.

That press conference is still the image people summon when the period comes up. If you were forty-five and running the treasury of a mid-sized manufacturer that autumn — Sanyo into court protection on the 3rd, Takushoku giving up on self-rehabilitation on the 17th, Yamaichi resolving to wind itself down on the 24th, twenty-one days end to end — the lesson on offer was that access to funding is held by somebody else. The three that failed were not themselves clean illustrations of it. Yamaichi had a quarter of a trillion yen (about $1.5 billion) it had not disclosed. Takushoku had given up on its own rehabilitation. The lesson was downstream of them, and it took its time.

The Shape of the Buildup

The buildup is visible in one series, though not as a reaction. Private nonfinancial corporations held 169.0 trillion yen ($1.03 trillion) of currency and deposits at the end of March 1998, and the level barely moved for the better part of a decade afterwards. They held 366.7 trillion yen ($2.24 trillion) at the end of March 2026, an all-time high, up 5.4% or 18.9 trillion yen ($115 billion) year on year, per the Bank of Japan's Flow of Funds Accounts published on 25 June 2026.

That is 2.17 times the 1998 level in nominal terms, unadjusted for prices or for the growth of the corporate sector, and measured on a series revised back to 1979 on the 2008 SNA basis — which is to say, not the number anyone was looking at in 1998. The latest figure is preliminary and will be revised.

It did not happen during the panic. The balance was 171.6 trillion yen ($1.05 trillion) at the end of fiscal 1990 and 183.9 trillion ($1.12 trillion) at the end of fiscal 2005: essentially flat, in nominal terms, across the crisis and its aftermath. What came later sped up rather than slowed — 193.4 trillion ($1.18 trillion) at end-March 2011, 239.2 trillion ($1.46 trillion) at end-March 2016, 279.1 trillion ($1.70 trillion) at end-March 2020, 366.7 trillion ($2.24 trillion) at end-March 2026. The bunker was not dug during the air raid. It was dug afterwards, most of it in the last fifteen years, fastest of all in the six years to March 2026, a stretch that included a pandemic.

Against nominal GDP of 669.4 trillion yen ($4.09 trillion) in fiscal 2025, itself a preliminary estimate, that comes to 54.8%, against 30.6% in fiscal 1997 — both endpoints sitting on retroactively revised bases. The pairing is two official series divided by each other, not an official statistic. The Cabinet Office publishes a corporate cash-to-GDP ratio of its own and put it nearer 60% at the end of 2023, on a wider definition of the corporate sector. The level moves with whose boundary you use. Either way it is a stock divided by a flow. It tells you how many months, not how much virtue.

The category is narrower than it looks. It covers currency and deposits and nothing else: 9.0 trillion yen ($55 billion) of cash, 254.9 trillion ($1.56 trillion) of demand deposits, 65.1 trillion ($397 billion) of time and savings deposits, 11.9 trillion ($72.6 billion) of certificates of deposit and 25.7 trillion ($157 billion) of foreign currency deposits, as of end-March 2026. Every security is excluded. And it is 20.0% of the 1,829.3 trillion yen ($11.2 trillion) of financial assets those companies hold at market value, alongside 680.0 trillion ($4.15 trillion) of equity and fund shares, against 2,746.4 trillion yen ($16.8 trillion) of liabilities.

A second official statistic counts differently. The Ministry of Finance's Financial Statements Statistics of Corporations puts cash and deposits at 267.4 trillion yen ($1.63 trillion) for all industries excluding finance and insurance, as of the quarter ended September 2025 — roughly 80 trillion yen ($490 billion) below the central bank's 347.8 trillion ($2.12 trillion) for end-March 2025, and close to 100 trillion ($610 billion) below its latest reading of 366.7 trillion ($2.24 trillion) for end-March 2026. No published decomposition of that gap turned up. At least three things drive part of it: the ministry's series is a sample survey that leaves out finance and insurance, it captures only cash and deposits booked as current assets, and the reference dates are six months apart.

The round figure that still circulates, 350 trillion ($2.14 trillion), is consistent with the same central-bank series one to two years earlier — 349.9 trillion at end-March 2024, 347.8 trillion at end-March 2025 — though which citations it descends from has not been pinned down.

A Request with No Deadline

On 31 March 2023 the Tokyo Stock Exchange sent every Prime and Standard company a document titled Action to Implement Management that is Conscious of Cost of Capital and Stock Price. It applies regardless of valuation; the Growth Market is not covered. It asks for a three-step cycle — analyze cost of capital and profitability at board level, disclose policies and targets, execute and update at least annually — and states that buybacks and dividends alone are not what is expected. It sets no deadline. It is not mandatory under exchange rules and carries no penalty. The exchange updated it in April 2026.

What it has instead is a list. From 15 January 2024 the exchange began publishing monthly the companies that have disclosed. Only disclosers appear. The compilation note is the whole design in one line: "If none of the keywords are found, the company will not be listed." No list of non-responders is published. The shorthand that travels in English, the shame list, is not the exchange's word.

It worked in the direction such things work, at least as the exchange measures it, which is a formal check of what the governance report says rather than of whether it says anything. As of 31 December 2023, 815 Prime companies had disclosed or were considering it, 49% of 1,656, along with 300 Standard companies, 19% of 1,619. As of 30 June 2026 it was 1,462 Prime companies, 94% of 1,552, and 877 Standard, 56% of 1,563. The endpoints are not quite like for like: from January 2025 the exchange began dropping companies that had sat at "under consideration" for more than six months. Of the companies on the list, 345 have asked to be flagged as wanting more contact from institutional investors.

The valuation statistic that made all this legible has moved a great deal. As of 1 July 2022, 50% of Prime Market companies traded below 1.0x book, 922 of 1,837, along with 64% of Standard and 7% of Growth; on an index basis, in data the exchange compiled from Bloomberg for its own follow-up council, it was 43% for TOPIX 500 against 5% for the S&P 500 and 24% for STOXX 600. As of 1 March 2026 it was 27% of Prime, 424 companies, and 49% of Standard — down 23 and 15 points.

Over the identical period the share of Prime companies earning less than 8% on equity fell from 47% to 43%, and Standard from 63% to 60%; on Standard the count of such companies actually rose, from 912 to 936, the share slipping because the market itself grew.

Price-to-book has a share price in its numerator. The 1 March 2026 reading was taken a month before the total value of domestic shares listed in Tokyo fell roughly 12%, from 1,372.4 trillion yen ($8.38 trillion) at end-February to 1,213.2 trillion ($7.41 trillion) at end-March. Return on equity has no price in it. One of the two ratios can be cleared by the market alone; the other requires the company to do something — and the something can be shrinking the equity base rather than earning more.

Buybacks, Dividends and Tallying Houses

The cash is nonetheless moving. Listed companies authorized a record 22.325 trillion yen ($136 billion) of buybacks in fiscal 2025, April 2025 to March 2026, up 18% and a fifth consecutive annual increase, on Nikkei's tally by announcement date. Count by calendar year, and from a different house — Daiwa Asset Management, working from QUICK data — and the same phenomenon reads as 17.8 trillion ($109 billion) in 2025 against 18.0 trillion ($110 billion) in 2024, a slight decline. Announcements from March-year-end companies bunch into April and May, 9.0 trillion ($55 billion) of them in 2025, and the pace fell away after July, so the total swings with where the year is cut and with who is cutting it.

On Daiwa's count of shares actually repurchased rather than merely authorized, 2025 was already 18% above all of 2024 by the end of November. Authorization, execution, fiscal convention and choice of tallying house give several stories about broadly the same activity, which is why the basis has to be named every time.

Dividends carry the same caveat. Nikkei's January 2026 tally projected total payouts to pass 20 trillion yen ($122 billion) for the first time in the year ending March 2026, up 8% and equal to about 40% of net profit across roughly 2,200 companies. That is a projection, and it has already moved once: in July 2025 the same tally read 19.99 trillion, up 3%, across about 2,300 companies. No confirmed full-year figure has been published.

And somebody is asking. A record 141 companies received shareholder proposals across the 2025 meeting season, 65 of them targeted by activist or institutional investors on Daiwa Institute of Research's count; activists filed 246 large-shareholding reports flagging important proposal activity in 2025, against 197 in 2024. Take June meetings alone, which is the like-for-like slice: 111 companies in 2025, a record, then 101 in 2026, the second-highest, while the activist subset set a new high at 52 against 51. The 2026 numbers are preliminary, as of 12 June.

The proposals mostly lose. Seven of the 111 companies saw one carry in June 2025, about 6%, and two did in June 2026. Board seats are the exception: Diligent Market Intelligence counts 37 won at Japanese companies in 2025 against 7 in 2024, and says most such seats globally arrive by settlement rather than a vote. Nobody has to win a proxy fight to be listened to.

The Stock and the Liabilities

None of this has teeth. The request carries no deadline and no penalty, and the list names only the companies that answered, including the ones still calling it under consideration. Nor is there a single number to argue over: the central bank's flow-of-funds accounts and the finance ministry's corporate statistics cover different populations, different balance-sheet lines and different dates, and in 2025 they stood about 80 trillion yen ($490 billion) apart — 347.8 trillion ($2.12 trillion) at end-March against 267.4 trillion ($1.63 trillion) at end-September — with the gap against the newest central-bank reading wider still.

What changed is quieter. That reading, still preliminary, puts corporate currency and deposits at 366.7 trillion yen ($2.24 trillion) at end-March 2026, up 18.9 trillion ($115 billion) in a year: the balance was still growing through the years in which the disclosure and valuation numbers improved. The same accounts carry 2,746.4 trillion yen ($16.8 trillion) of liabilities, so the stock on its own settles nothing. The cohort that was forty-five in November 1997 is seventy-three or seventy-four now, and by arithmetic alone the decisions are passing, one board meeting at a time, to people for whom that press conference is archive footage.

Every yen figure here is converted at 163.8 to the dollar, the rate on 24 July 2026, including the ones from the 1990s, which nobody was converting at anything like that rate at the time. Comparing the dollar figures across years therefore tells you nothing about the exchange rate; it only restates the yen. The yen numbers are the record, and the dollars are an approximation for the reader's convenience.