A Blueprint for the Waterwheel

First, clear up a misconception. Japan's surplus does not come from exports. Of its JPY 32.2 trillion current-account surplus in 2025, the trade balance was −JPY 0.6 trillion and services were −JPY 3.4 trillion — goods and services ran a deficit. All of the surplus came from primary income of +JPY 41.8 trillion: dividends, interest and retained earnings from overseas assets. The latest 12 months in dollars tell the same story: goods plus services −$13.7bn, primary income +$269.9bn.

The turning point was 2005, when primary income first overtook the trade balance. The gap has widened in one direction since. In 2022, trade fell to −JPY 15.5 trillion while primary income earned JPY 35.0 trillion. Japan is a creditor nation rather than a trading nation. The water feeding the wheel comes from overseas balance sheets, not factories.

This composition reverses what a weak yen means. Over 30 years, the yen exchange rate and the trade balance are uncorrelated (−0.05), while the correlation is +0.53 with primary income, rising to +0.95 from 2011 onward. A cheaper yen improves the yen value of overseas earnings rather than export competitiveness. Yen weakness works as a currency-translation mechanism, adding more water to the wheel.

In #11, I wrote that Japan's portfolio investment had turned to net inflows (+$19.2bn). Money is still leaving Japan, however. Split the financial account into layers: portfolio investment (L1) shows +$188bn of net inflows, but the corporate and institutional layer (L2), combining direct investment, other investment and derivatives, overwhelms it with −$443bn of net outflows. The total (L3) is −$255bn. Individuals are not the force behind the weak yen; it is corporate capital exports, rather than individuals.

This is the reverse of Korea. Korean households drive outflows through securities accounts (retail investors buying overseas stocks — offset ratio 44%). Japanese companies do it through factories and acquisitions. Their overseas earnings return as primary income (§01), and some go to domestic shareholders (§03). Companies earn abroad and pay shareholders at home — that cycle is the waterwheel's skeleton.

Two caveats belong here. First, portfolio investment also had −$10.5bn of outflows in the latest 3 months (May–July). The +$188bn is a legacy of the preceding 9 months; its direction is being tested again. Second, these inflows themselves are a function of the US–Japan interest-rate gap. As the gap narrows, the reason to keep money abroad weakens and money returns. The waterwheel's rotation rests on the rate environment. This sentence returns in §06.

Payouts can be measured. Listed Japanese companies report their buybacks to EDINET each month. Add up those 5,507 reports: the latest 11 months (TTM) total JPY 18.51 trillion, or an annualized 1.696% of market capitalization. Add these observed buybacks to reported dividends (a 1.94% lower bound) and Japan's total payout doubles. More on that in §07. First, who is buying?

The lower ranks matter more. Over 11 months, Bridgestone bought in 10 months. Kyocera, ORIX, Sompo and Advantest bought in 9 months; Mizuho, MS&AD, Japan Post, Hitachi and Recruit in 8 — monthly buying has become routine. Mizuho's 8 consecutive months are especially telling. A bank once late to shareholder payouts has become a habitual buyer.

Compress this landscape into one chart and a rule appears: the high ground of payouts is the low ground of P/B. Toyota at 0.95, Denso at 0.95 and Japan Post at 0.68 pay out 6–12%, while Recruit (P/B 13.0) pays just 1.7% and Advantest (24.3) just 0.7%. Even in Japan, companies that can sell a growth story cut off the fruit — the proposition from #10 reappears within Japan. Higher payouts are not universal across Japan; they are a phenomenon of old-economy, financial and low-P/B companies.

JPY 4.9 trillion, 26% of JPY 18.51 trillion, is different. It was a chain of TOBs last spring around Toyota Industries' going-private transaction. Each group company bought back its shares held by Industries through an own-share tender offer and canceled them: Toyota JPY 3.66 trillion (all of Industries' 9.15% stake), Toyota Tsusho JPY 663.7 billion and Denso JPY 313.6 billion. Kansai Electric used the same approach to sell Kinden shares (JPY 223.7 billion, 37.13%→24.33%). This chain unwound cross-shareholdings.

Why an own-share TOB? Sellers accepted a below-market price(Kinden −11%, Toyota −3%) because of taxes. When a corporation tenders into an own-share TOB, much of the proceeds is treated as a deemed dividend, eligible for the corporate dividend exclusion from taxable income. After-tax proceeds can exceed a market sale, where the full capital gain is taxed. The seller gains through taxes, the company through capital efficiency, and remaining shareholders through the cancellation of cheaply bought shares: a transaction designed to benefit all 3 parties.

This yields the report's first conclusion. Japan turned the water back through institutions, not corporate goodwill. The exchange pressures low-P/B companies (the §03 map), while tax rules provide a buyback exit from cross-shareholdings (the §04 surgery). Standards and tax law cut the waterwheel's teeth.

Kinden is a case of rising water. The Kansai Electric-affiliated electrical contractor's order backlog reached a record JPY 581.8 billion (+23.2%) amid a data-center and chip-factory construction rush. Net income rose +47% in a year to JPY 69.4 billion. As the US draws in global capital with the AI story, that capex lands in Japanese electrical-construction order books. The surging profits reached domestic shareholders through a TOB (JPY 223.7 billion) and a special dividend (JPY 100 of the annual JPY 240). Annual payouts of JPY 263.2 billion — 3.8 years of net income — were funded with accumulated cash and JPY 230 billion borrowed from Mizuho. This company's books show where the AI funding channel ends in Japan.

Toyota is a case of falling water. In the year it became the first Japanese company to top JPY 50 trillion in revenue, net income fell −19.2% to JPY 3.85 trillion. Next year's guidance incorporates an annual tariff hit of −JPY 1.45 trillion, leaving JPY 3.0 trillion (−22%). In that year Toyota executed the Industries-related JPY 3.66 trillion TOB, then on August 4 announced a new buyback ceiling of JPY 1 trillion (4.22%) + cancellation of 200 million shares. Do the math: combine earnings down −22% with fewer shares and the EPS decline compresses to around −10%. The dividend also rises from JPY 95 to JPY 100. Payouts are designed as a cushion against the tariff shock, rather than a growth story.

Together these cases show the waterwheel's condition: where water rises (AI capex), payouts grow with profits; where water falls (tariffs), payouts defend per-share value. Either way, the wheel turns — so far.

Here we must identify the hidden premise. Kinden borrowed to buy back shares. Toyota's new JPY 1 trillion and Mizuho's 8 consecutive months of purchases all rely on a world where the cost of capital is below the payout yield. That premise has begun to crumble: JGB 30Y 4.002% (8/13), 10Y 2.873%. The 30-year's 4% level has been reached. For companies with total payout yields around 4–5%, the margin on borrowing to buy shares is now paper-thin.

The sentence from §02 returns here. Net inflows of +$188bn rest on the US–Japan rate gap, which has narrowed to 1.757%p — squeezed by rising Japanese rates and falling US rates. The US 30-year stands at 5.21%, up +13bp in 21 trading days. That rise comes from Real yields, rather than inflation expectations. Japanese long yields rise alongside US yields, but the US slope is 4x steeper. The waterwheel turns, yet the price of water is rising on both sides. Turning the water back is a product of institutions and of low rates. That is this report's second conclusion.

Now add the §03 measurements to reported figures. Japan's reported total payout yield of 1.94% is a lower bound that omits buybacks. Add observed EDINET buybacks (annualized 1.696%) and it becomes 3.64% — above the US at 3.06%. Even excluding Toyota's one-off TOB, it is 3.30%, still higher. In #08, I wrote that Japan had caught up. On an observed basis, it may already have overtaken the US.

The reasons for caution are explicit. ETF distributions understate dividends (a lower bound); buybacks and dividends have different denominators; and US calendar-year data and Japanese TTM data have different windows. The dotted line is therefore an interim step toward measurement rather than a settled measurement. Next summer, after 12 months of monthly EDINET accumulation, we will confirm it with our own annual series. Until then, the overtake remains a candidate finding.

Turn the mirror to Korea. On the §03 map, low P/B meant high payouts — in Japan. At similarly low P/B, Korea's total payout is 1.60%, last among 5 markets. Water also kept draining from its stock market through the second week of August. Over 20 trading days, foreign investors withdrew from memory semiconductors KRW -6.37 trillion; biotech received +KRW 0.29 trillion — 5% of the money withdrawn. The rest left the Korean market itself. Rather than rotation, this was an exit.

Just before this report closed, a test result arrived. Over the three days of 8/12–14, foreign investors bought back in the KOSPI KRW +7.99 trillion(+2.84 → +2.11 → +KRW 3.04 trillion — accelerating). The index rose +10.0% over three days and touched 7,000 intraday. The memory outflow gap rewound from KRW 6.66 trillion to KRW 4.01 trillion. Of three spikes since late July, the first two reversed within three days. This was the first to hold. Three caveats keep judgment open: foreigners still sell KOSDAQ stocks (a narrow preference); the change of hands absorbed −KRW 7.9 trillion sold by individuals over three days; and the return came during three days of respite in US yields(30Y 5.24→5.21%). If rates set the direction, that KRW 8 trillion could leave just as fast when yields rise again.

What the 3rd Commercial Act amendment should learn is the mechanism in §04, rather than a campaign. Japan's buyback surge came through an exit built by tax rules (dividend income exclusions) and exchange standards (pressure on low P/B), rather than a slogan to pay shareholders. Without a structure benefiting the seller, company and remaining shareholders together, Korea's cross-shareholdings and low valuations have no reason to unwind. Draw a waterwheel without building its machine, and the water keeps flowing to sea.

First, settle the #11 gauge. Item 1, “persistence of directional KOSDAQ capital,” was disproved in a day: on 8/11 the institutional directional share fell 44%→22%, and foreigners turned sellers. Gauges exist for this purpose, so we record it as it is. Keep the other needles and hang seven new ones for this report.