Two Illusions
The first mistakes two different metrics for the same number; the second confuses the cause with the amplifier. Beneath the visible flows, we read the submerged structure.
Selling,
Yet Owning More
From the start of the year through July 3, foreign investors net sold KRW 150.2627tn in the KOSPI. Of that, Samsung Electronics and SK Hynix accounted for KRW 136.0444tn, or 90.5%. Reports citing early-July KRX data put the foreign ownership share at 41.58% according to the tally.
More precisely, it was not the number of shares held, but the market-wide foreign ownership share by value that increased.
| Category | Since Year-Start |
|---|---|
| Net Selling | KRW 150.2627tn · Record First-Half Outflow Samsung Electronics KRW 75.5135tn · SK Hynix KRW 60.5309tn · Combined KRW 136.0444tn, 90.5% |
| Samsung Electronics Foreign Ownership | 52.33% → 46.55% · 17-Year Low |
| SK Hynix Foreign Ownership | 53.83% → 49.87% · 3-Year, 2-Month Low |
| Market-Wide Foreign Ownership Share | 36.65% → 41.58% · Reports Citing Early-July KRX Data |
Based on reports citing KRX data · Net selling: 2026.01.02–07.03 · The 41.58% foreign ownership share is based on an early-July snapshot.
Foreigners reduced their share counts in both stocks. Yet the prices and KOSPI weights of the shares they still held rose even faster. A rising ownership share does not prove fresh accumulation; it can be the result of market-cap-weighted valuation. Company-level foreign ownership counts shares; market-wide foreign ownership measures value. The KOSPI rose from 4,214 at the start of the year to an intraday 9,385 on June 19 — +122.7% in total.
the statement that foreign ownership is at a year-to-date high
can both be true.
This is where the case for further selling comes from. As valuations rise and the weight increases, Korea's share in a portfolio can exceed its target, creating additional rebalancing pressure. KB Securities translated this into a notional amount. If the foreign ownership share were reduced from 39.5% to 35%, potential additional selling would be about KRW 260tn. This is a scenario calculation, not an actual selling forecast.
But a high weight does not automatically foreshadow selling. Actual motives can include profit-taking · country and sector rebalancing · FX moves · earnings downgrades. In a July 20 note, Citi instead argued that flow headwinds are easing and characterized the recent correction as rebalancing and profit-taking. The same flow data can support different directional views.
Retail's Choice,
the Footprint Left in Institutions
The difference between the table that includes ETFs and the one that excludes them is the ETF component. The question is where retail's choice is recorded and that is where the split appears.
| Investor Type | Ex-ETF | Incl. ETF | Difference · ETF |
|---|---|---|---|
| Retail | +KRW 17.76tn | +KRW 27.33tn | +KRW 9.57tn |
| Foreign | −KRW 17.91tn | −KRW 16.78tn | +KRW 1.13tn |
| Financial Investment | +KRW 1.70tn | −KRW 12.60tn | −KRW 14.30tn |
| Institutions, Total | −KRW 0.64tn | −KRW 11.89tn | −KRW 11.25tn |
KRX investor trading data · Total market · Cumulative net buying, 2026.07.01–28 · ETF component calculated as the ETF-included table minus the ETF-excluded table. The two tables use identical query conditions except for ETF inclusion.
In July Retail bought KRW 9.57tn in ETFs alone. On the other side, Financial Investment was the largest supplier, net selling KRW 14.30tn. LP/AP activity was likely a substantial part of this, but the aggregate table alone cannot establish that all of it was liquidity-provision trading.
From launch through June 19, retail investors net bought KRW 8.20tn of single-stock leveraged products. Over the same period, retail average daily net buying in ETFs excluding single-stock products fell from KRW 551.9bn before launch to KRW 116.4bn in the first week after launch; domestic semiconductor ETFs swung from +KRW 171.7bn to −KRW 286.6bn. A meaningful portion may have reflected migration of existing ETF money rather than purely fresh capital.
AUM growth is not the same as net creations. AUM in SK Hynix leveraged ETFs rose from KRW 4.84tn on June 10 to KRW 9.15tn on June 19, an increase of KRW 4.31tn, but researchers estimate roughly KRW 3.60tn of that came from NAV appreciation as the stock rose. Retail investors actually net sold KRW 140bn over the same period.
※ To determine how much retail ETF buying translated into new share creation and underlying-asset hedging, check shares outstanding and net creations.
but it leaves two records.
institutions made a discretionary call.
If the aggregate institutional bucket is read as a single investment decision, the interpretation breaks down. Financial Investment activity includes ETF hedging and arbitrage, while pension funds and insurers behave differently. July aggregates alone cannot isolate how much of Financial Investment net selling was creation/hedging linked to retail ETF demand. In the total market, pension funds were nearly flat, with KRW 129.1bn of net selling ex-ETF in July. They simply sit inside the same 'institutional' bucket.
Whether retail's KRW 9.57tn of ETF buying flowed through new creations into the underlying assets or merely changed hands in existing units depends on shares outstanding and net creations. This report did not obtain those data. We map the route, but do not claim a verdict.
The Hand That Lifts and
the Hand That Catches
In July, excluding ETFs across the total market, retail was +KRW 17.76tn, while foreign investors were −KRW 17.91tn. The two almost offset one-for-one. Yet the index fell.
Figures are based on the KRX total market · they do not map one-for-one to the KOSPI index.
There is no contradiction in the index falling 5% on a day when retail net buys KRW 1tn. Net buying tells us how much was bought, not the prices at which it was bought. And prices are set by the side in greater urgency.
Buying continues even as prices rise. Foreign supply is absorbed at successively higher execution prices.
Retail net buying increases as prices fall, absorbing selling into weakness.
As a result, domestic capital can be interpreted as having absorbed foreign selling at progressively higher prices. If buyers were merely waiting at lower prices, it would be difficult to see heavy foreign net selling alongside a surging index. Margin financing rose as well, reinforcing chase demand.
After the first week post-launch, retail showed a contrarian pattern: buying more as prices fell and less as they rose. This was different from buying that chased prices higher. Public aggregates cannot tell us whether investors crossed the offer or waited with lower limit orders.
became a market that fell roughly in line with the selling.
the disappearance of chase buying
And then, on July 29, even the catching hand stepped back. Retail turned seller for the first time in this phase, with KRW 1.9701tn of net selling.
Then why did the entire index shake so violently once chase buying disappeared? The answer lies inside the index. The two stocks bearing the concentrated selling pressure had already become half the index.
Margin financing based on reports as of Jul. 6 · About KRW 9tn of new creations is a Citigroup estimate · Retail's contrarian pattern from Korea Capital Market Institute analysis of 2026.06.04–19 · Jul. 29 flows are final KRX data · The 'lifting hand / catching hand' distinction is an interpretation combining flows and price response, not an order-book classification.
When Two Stocks Became
Half the Index
The line below shows the place occupied by Samsung Electronics and SK Hynix in KOSPI market capitalization. Their combined weight rose from 34% at end-2025 to 41% at end-April, 49% on May 26 and 52% on July 15.
Combined · 48%
This picture says something simple. If you tracked the market-cap-weighted KOSPI, nearly half your money was effectively tied to two stocks. You were buying something closer to two semiconductor stocks than the index. Their news therefore became the index's news.
On top of an already concentrated pair, on May 27 single-stock leveraged products were added. Fourteen leveraged and two inverse products—16 in total—grew from KRW 4.4tn in market cap on listing day to KRW 11.9tn by July 15, nearly tripling in seven weeks. Trading value reached KRW 13tn, accounting for 38.2% of all ETF trading value. The two trends grew together, but that alone does not prove the products created the concentration.
Korea Capital Market Institute data put cumulative net buying of leveraged products at KRW 8.2tn (SK Hynix KRW 4.6tn · Samsung Electronics KRW 3.7tn). As of June 19, AUM was KRW 9.15tn for SK Hynix products and KRW 5.22tn for Samsung Electronics products.
Once the index became two stocks,
the earnings dates of those two stocks
became judgment day for the index.
The same path worked in reverse on the way down. At 1:10 p.m. on July 29, the two stocks' combined weight fell to 47.3%. The concentration had not disappeared; the concentrated area had simply been marked down. Regulation followed. New listings and advertising were suspended from July 16. From July 31, a KRW 30mn cash minimum deposit will apply, with substitute securities not accepted. Tighter tracking-error controls take effect Aug. 19; a larger trading unit of 20 units is targeted for November, subject to consultation. At a July 28 meeting, regulators said that if the measures proved insufficient they would consider per-investor investment limits (for example, within 20% of total invested capital) and retraining / simulated trading, and asked the industry to distribute rebalancing intraday rather than bunching it near the close.
Market-cap weights and product figures: Financial Services Commission supplementary measures, 2026.07.16, and early-implementation notice, 07.24 · Jul. 29 47.3% as of 1:10 p.m. · Cumulative net buying / AUM: Korea Capital Market Institute.
Is Leverage
the Culprit?
The most widespread explanation is that leveraged products created the depth of the decline. The numbers tell a somewhat different story.
The two numbers measure different things. One estimates the degree of adjustment in leveraged ETF exposure; the other the share of margin positions liquidated. Neither is exchange data; both are model estimates by the respective firms.
had already declined substantially.
Yet on July 29 the KOSPI
plunged even after substantial deleveraging had already occurred. Deleveraging alone is insufficient to explain the full decline. Remaining exposure is still not small, however, and in thinner liquidity the price impact of the final liquidation can exceed earlier rounds.
The second number is more decisive. Across the period when single-stock leveraged products were traded, May 27 through June 19, putting Korea and U.S. semiconductor volatility side by side gives the following picture.
Analysis Period: 2026.05.27–06.19 · Annualized volatility; bar length = increase in percentage points · The comparison covers the first three weeks after launch, not the July selloff · Korea Capital Market Institute.
Volatility did not rise only in Korea. Over the same period, volatility increased even more in global semiconductors where no such product existed. That means the cause cannot be located solely in domestic leveraged products; the Korea Capital Market Institute also interprets the move as the overlap of multiple factors. The institute itself explicitly notes that a sample shorter than one month after launch is too short for rigorous causal inference and that AUM includes LP holdings, so not all of it translates into market impact.
To maintain the daily leverage multiple, same-direction rebalancing does exist. But Korea Capital Market Institute estimates that cash-equity rebalancing averaged 1.6% (Samsung Electronics)–2.1% (SK Hynix) of trading value, generally 1–3% for Samsung Electronics and 1–4% for SK Hynix. For SK Hynix on June 19, additional buying was estimated at about KRW 260bn in cash equities and KRW 270bn in futures. The average scale alone is insufficient to explain the full selloff, but when liquidity thins and trading clusters near the close, it remains a channel that can amplify an existing price move. This is also why the Financial Services Commission asked the industry on July 28 to spread rebalancing through the session.
the products were layered on top.
Deleveraging has not moved in one direction either. Margin financing rebounded to KRW 33.194tn on July 28, then fell by KRW 199bn to KRW 32.995tn on July 29. Deleveraging did not end in a straight line, nor did new leverage build in only one direction.
The bill was large. The figures below are model-based estimates from Citi trading strategists.
FX rates, cut-off dates and product universes differ, so these figures are not a simple accounting identity to add and subtract. Citi's universe appears broader than the 16 domestically listed single-stock products, and the figures are model estimates, not confirmed realized losses.
Cited Material · Reference Dates
- Korea Exchange — investor trading data (total market, with and without ETFs), cumulative 2026.07.01–28. ETF component calculated as the difference between the two tables.
- Korea Exchange — investor-by-stock flows and final closing prices, 2026.07.29.
- Financial Services Commission — ‘Supplementary Measures for Single-Stock Leveraged Products’ 2026.07.16 / ‘Early Implementation of Strengthened Minimum Deposit Requirement’ 2026.07.24 / July 28 meeting: additional measures under review (investment limits; request to spread rebalancing intraday).
- Korea Capital Market Institute — ‘Stock-Market Trends and Implications Before and After the Launch of Single-Stock Leveraged/Inverse ETFs,’ 2026 — retail net buying, existing ETF flows, AUM/NAV, contrarian trading, and rebalancing estimates for May 27–June 19.
- Federal Reserve, Are Leveraged and Inverse ETFs the New Portfolio Insurers? / Are Concerns About Leveraged ETFs Overblown?
- Citigroup Global Markets Securities, Mohammed Apabhai — 65% of YTD margin positions liquidated (as of Jul. 28); estimates of additional inflows and losses during the decline. JPMorgan — estimated Korea-related leveraged ETF deleveraging about 75% complete (published Jul. 21).
- Citi — KOSPI 12-month forward P/E 5.8x (as of Jul. 16), below 6.4x in 2008 and 8.5x in 2020. Target index and upside from Jul. 20 note.
- KB Securities — estimated potential additional selling if foreign ownership share falls from 39.5% to 35%, Jul. 6.
- Foreign first-half net selling, stock-level ownership ratios and market-wide ownership share cross-checked across reports citing KRX data (Financial News · Herald Economy · Aju Business Daily).
Reference dates — domestic index and flow data are final as of the Jul. 29 close. The two stocks' combined 47.3% market-cap weight is an intraday reading at 1:10 p.m. on Jul. 29. First-half figures cover Jan–Jun. The FOMC result released early Jul. 30 is not reflected. Brokerage estimates are shown with source and reference date and are not the editorial team's forecasts. This report is an explanatory market-structure note, not a recommendation to buy or sell any security.