The Value of Savings
Foreigners sold KRW 165 trillion in Korean stocks this year. Yet the index rose. The other side of #16's unanswered question is now clear: who bought those shares. Of the buying side, 61% was retail investors and 15% was the companies themselves (share buybacks). The national pension fund was selling on the other side, driven by a formula rather than intent. In an era of managed prices, savings are the last buyer. The value of savings becomes the value of a country's stock market.
The Buying Hands
Pensions Move Against the Flow
Pension funds have net sold KRW 8.9 trillion this year (consistent with reported January–August net selling of KRW 8.8 trillion: cross-validation). It looks like Value-up (Korea's corporate value-up program) and its second paradox: the state calls money into stocks, while its pension fund leaves. But the mechanism is a formula. The NPS's target domestic-equity allocation was actually raised this year, from 14.9 to 20.8%. The problem is that the index rose faster. The actual allocation reached 29.1% (officially reported at end-June), leaving about KRW 160 trillion of excess holdings. Rebalancing rules now require selling.
Even using the full permitted range (SAA ±6pp), KRW 51–57 trillion in potential sales remain, according to brokerage estimates★. The mechanism is simple: The higher the index rises, the more pension funds must sell. A managed target-allocation formula becomes an automatic selling mechanism in a rising market. This overhang above KRW 102.0 trillion in retail net buying is one vulnerability in §05. ("Pension funds etc." is a KRX category, not the NPS alone. Its definition differs from the fund's own disclosures; see the footnote.)
Mobilization in Two Countries — Real Buying and an Illusion
The Last Buyer
This joins #16 and #17. The Commercial Act ordered cancellation (#16), and P/B reform pressed shareholder returns. In both Korea and Japan, the last buyer created by the rules was the company itself. Price governance (institutional instructions) and mobilized savings (households entering) meet at share buybacks. In an era of managed prices, the last buyer is the company itself, instructed to raise its value.
| Case | Details |
|---|---|
| SK D&D — A rights offering cleared by lowering its price | It answered the correction request (8/5) with a lower price: stage 1 issue price, 3,060 →KRW 2.26 thousand (18% of the KRW 12,750 tender price). The 44.68 million new shares (240% of existing shares) stayed unchanged; proceeds fell from KRW 136.7 billion to KRW 100.9 billion. The collapse cut the controlling shareholder's allocation cost (KRW 80.1 billion) by 26%. Effective 8/28 → ex-rights 9/1 (reference price 2,810) → rights certificates list 9/21 → final issue price 10/6 (lower of stage 1 and stage 2) → subscription 10/12 → payment 10/15 → new shares list 10/28. Shinsegae Food answered 2 correction orders with a higher price (appraisal rights exercised for 580,111 shares; 55.66% of minority shareholders exited for cash). SK D&D lowered its price and cleared: the same tool, opposite responses. Once effective, authorities have only Article 131, premised on false statements |
| SK D&D — Two days of illusion; price closes the door | 9/1 upper limit 3,650 → 9/2 4,220 (+15.6%, surrendering the intraday upper limit; volume 6.24 million shares = 1.6x the 3.84 million-share free float). Up +50% in two days from the reference price. Subtract the cash raised from fully diluted market capitalization, and the existing business is valued at 2.2x its value three days earlier, without new information. That price does two things. If unallocated shares are not issued and the market price exceeds the issue price, minorities subscribe. If they subscribe, the controlling shareholder's stake (79.36%) does not rise. The rebound is fuel for the rights offering and a brake on ownership concentration. New buyers without rights pay that price. An 8/31 holder exits with old shares at 4,220 + 2.4 rights certificates at 4,704 =8,924 (70% of the tender price they rejected). With neither certificates (9/21) nor stock borrowing, the half-market has no tool to lower the price. It continues until certificates list on 9/21 and new shares on 10/28 (9.22 million shares costing 2,260) |
| SK D&D — Price, not ownership, is the barrier | The delisting threshold is not 95%. A comprehensive share exchange needs only a shareholders' special resolution (2/3 of votes present and 1/3 of shares outstanding), and the consideration can be cash. LocknLock, Connectwave, Ecomarketing and Viol exited through this door; Hahn & Company had the votes before its tender offer. The barrier was the pricing rule, which is changing. The Capital Markets Act amendment passed by the National Assembly on 8/20 replaces market prices in affiliate mergers, share exchanges and appraisal claims with fair value (combining market, asset and earnings values). Effective 3 months after promulgation, applying to board resolutions after that date★. Including asset value for low-P/B companies can put exchange prices well above market prices. The period from 10/28 until implementation is therefore the last window under market-price rules. Disclosures will reveal intent: controlling-shareholder certificate sales on 9/21–28, subscription results on 10/12–13 and board resolutions within that window. The filing says the largest shareholder plans to subscribe for its entire allocation (without stating the funding method). Selling certificates would conflict with that statement |
| Samsung — Nine decimal places, and two days of buyers | On 8/21, Samsung Electronics disclosed KRW 90–110 trillion of shareholder returns. The original specifies only KRW 30 trillion entirely in cash dividends in Q3; the end-October/end-January boards will decide cancellation amounts. The same day's KRW 15 trillion buyback is for compensation and will not be canceled; acquisition ends on 11/21. The market read the arithmetic, not the announcement. Cancellation raises remaining shareholders' ownership percentages. Samsung Life and Samsung Fire's combined stake in Samsung Electronics was 9.999999986% ★ (reported), close to the 10% financial-industry ownership limit. Cancellation → higher ownership → limit → selling already played out in March's KRW 1.4–1.5 trillion block deal ("preemptive response to the financial-industry law"). On 8/24, Electronics −8.70% (257,000) and Life −13.09% fit that configuration (#16). From #17, another piece appears. In KOSPI's 9/1 and 9/2 sell-offs, other corporations alone absorbed KRW 1.6657 trillion and KRW 1.6505 trillion, respectively: the category recording direct buybacks (only a consistent configuration until stock-level decomposition). Even as the instructed buyer bought, Samsung Electronics on 9/2 was 250,500 (−4.02%), 11% below the 8/21 close of 281,500. The end-October board will set the first cancellation increment, which sets Life and Fire's next sales. There is no basis to say the returns were designed around the limit. We can say only that two laws squeeze the same company from opposite sides |
The Buyer's Vulnerabilities — Korea
| Axis | Current | On 7/28 | Reading |
|---|---|---|---|
| Leverage intensity | 0.852 | 0.98 | Rebuilding — still below |
| Semiconductor concentration | 42.5% | 60.8% (trading-value peak) | Eased, but still high |
| Retail concentration | July net buying KRW 5.0 trillion → August KRW 5.4 trillion | — | Similar level persists |
First, Pension overhang (§02): a selling instruction that grows as the index rises hangs above retail buying. Second, concentration by market. Retail net buying of KRW 102.0 trillion is a KOSPI story (KRW 109.5 trillion). In KOSDAQ, retail investors were net sellers of KRW 7.6 trillion, while foreigners net bought KRW 4.1 trillion. The flow pattern is reversed. Mobilization was KOSPI's mass migration into semiconductors. The map of buying hands is really a KOSPI map. Mania is the vulnerability in a retail-supported market. The three axes above form its gauge panel, measuring rather than forecasting.
Gauges — v7
This table is the gauge panel above the title. Like #13–#16, it is a live widget that updates with new observations after publication. Publication values remain in this table. The first five slots are Korea: mobilization gauge, pension overhang, leverage intensity, cancellation decisions and won–memory co-movement. The last three track savings crossing borders: real−BEI, ERP + cracks, and the Yen switch (a new needle showing the L2 sign alongside the US–Japan gap). Cumulative foreign flows were absorbed into ①, and the bill share lost its place among the three US slots. These needles were removed for this edition, not retired across all reports. The #15 and #16 gauge panels keep updating those needles live.
| # | needle | Now (at publication) | How to read it · If it moves |
|---|---|---|---|
| ① | Mobilization gauge | +KRW 127.0 trillion (retail 102.0 · other corporations 25.0) · 76% of foreign selling [Trigger · 09.02] | Retail investors + other corporations' cumulative 2026 buying:≥KRW 50 trillion AND absorbing ≥50% of foreign net selling → trigger; ≥KRW 20 trillion → watch. ★'"Absorbed"' means only a a consistent alignment. These data cannot establish persistence or intent. KRW execution basis cannot be compared with the balance of payments; ETF channels cannot be separated (source limitation). Other corporations include ordinary corporate trading, not only share buybacks. |
| ② | Pension overhang | Net selling KRW 8.9 trillion (cumulative · KOSPI 8.5) · allocation 29.1% vs target 20.8% (end-June) [Watch · 09.02] | Pension funds etc.' cumulative 2026 net selling:≥KRW 15 trillion → trigger; ≥KRW 5 trillion → watch. ★'Pension funds etc.' includes mutual-aid associations and other pensions, not the NPS alone. It differs from the reported'"NPS −KRW 8.8 trillion"'. Close readings provide cross-validation, not an identical series. Rebalancing overhang is KOSPI only (effectively 0 in KOSDAQ). The allocation is officially reported by the fund (29.1% as of 2026-06-30); the 20.8% target was approved by the fund committee on 5/28. Parentheses on the value line give the As of, or'(reported)' if the source is unconfirmed. The reported'~30%' estimates the later index surge and has no reference date. Only observed cumulative pension-fund flows determine the state; the monthly, low-frequency allocation does not change its color. |
| ③ | Leverage intensity | 0.852 (09/01) · −13% vs 7/28's 0.98 [Wait · 09.01] | KOSDAQ margin loans ÷ (index/100). A return to 0.98 (7/28 sell-off) plus a credit-driven increase → trigger; ≥0.88 (2019 trade-dispute low: 0.878) → watch. A falling index alone raises intensity, so Δ decomposition (credit/index contributions) filters that illusion. The standard denominator is market capitalization; without that series, index/100p substitutes. The conclusion that leverage is 'light' is therefore conservative. Credit is published a day later than the index, so their as-of dates differ. No composite mania score: crowding and concentration appear separately in the body table. |
| ④ | Cancellation decisions | 2026: 420 decisions · monthly average 46.7 = 2.8x 2024 [Watch · 09] | Versus the base year's monthly average:≥3x → trigger; ≥2x → watch. Decision disclosures lag execution (acquisition usually takes 3 months). A 'decision this month' does not mean 'buying this month'. Counts only, not amounts. |
| ⑤ | Won–memory co-movement | Joint strength · unit price +6.4%MoM / won −4.6%20d [Wait · 09.02] | 4 quadrants: co-movement (strength/weakness) fits the assessment → wait; the 2 divergence cases → watch; assessment falsification condition 1 ('continued strength despite falling unit prices') → trigger. Mixed frequency: monthly unit prices × daily exchange rates. Outside a unit-price update month, only the exchange rate has moved. |
| ⑥ | Real−BEI divergence | Real 30y +6bp / BEI +3bp (5 days) [Wait · 09.01] | BEI 5-day change |10bp| or more → trigger (intervention reaches expectations); divergence |5bp| or more → watch. While only real yields move, this is the fingerprint of supply-and-demand intervention. |
| ⑦ | ERP + cracks | −1.39pp · B200/H100 2.24 · Chinese providers 54.7% · cracks 08/31 [Watch · 09.02] | 21-day change ≤−30bp → trigger; negative ERP → watch (speed becomes valid after enough observations). ERP is daily; the crack signal is weekly. The chip date refers to ERP; the crack date appears in the value. Do not read them as same-date values. While 13-week/4-week changes are None, accumulation is insufficient: do not infer a direction. |
| ⑧ | Yen switch | US–Japan gap 180bp · L2 −$443bn · condition unmet [Wait · 09.02] | L2 (corporate/institutional direct and other investment, 12-month net outflows) turning positive → trigger. This is the switch, not the interest-rate gap. A US–Japan gap ≤170bp means watch only: carry-unwind pressure, not the main flow. Cyclical-to-structural = 1:89. The components differ (stock vs 12-month flow); do not interpret them as 'a multiple of relative importance'. Read only the difference in scale. Clear values if the source is more than 21 days old. |
The Value of Savings
Savings have long been the market's last safety net, a place to turn when markets collapse. In Korea now, the reverse is happening. Savings have stepped forward as the market's The Last Buyer. Retail investors entered directly (KRW 102.0 trillion net buying), companies bought their own shares (KRW 25.0 trillion), and professionals sold. Pension funds stood on the other side under a formula's instructions.
The buyer's identity redraws the map of vulnerabilities. Mania is the weakness of a retail-supported market. This edition's gauge needles therefore measure vulnerabilities with its needles rather than predict. In cases such as the pension overhang, the management formula itself becomes an automatic selling mechanism. In an era of managed prices, savings have become the value supporting prices. The value of savings is the value of a country's stock market.
Stories remain untold: the financial vulnerability of the buybacks supporting Japan's rally; the decomposition of "excess yen supply" (carry was not the main component); and what savings buy when they cross borders. Next time, we will cross borders. Meanwhile, dates to watch: 9/16~17 (FOMC: as of 9/1, markets priced a rate increase at roughly 2/3 probability) →9/21 (Vietnam upgrade; SK D&D rights certificates list) →10/6~15 (SK D&D final issue price, subscription and payment) →end-October (Samsung's board). Autumn's calendar will write the next chapter.