If you learn to read one daily data release out of Taipei, make it the three institutional investors report. The practical rule most local desks work by: foreign investors set the market’s trend, investment trusts set the sector rotation, and dealers are mostly noise — and a single day’s print means far less than a streak, especially once you cross-check it against futures positioning. This guide explains how to do exactly that, using real published figures.
Interest in this dataset spiked in late July 2026, when a two-session TAIEX rebound (the index rose 0.26% on July 23, 2026, its second consecutive gain) was explicitly attributed to institutional buying — including domestic mutual funds — after heavy retail margin unwinding on July 17. Foreign institutional capital represents roughly 40% of Taiwan’s market, so when commentators say institutions drove a rally, this report is how you verify it. Everything below is timeless mechanics; all figures are date-stamped.
Who the three institutions are
Taiwan’s exchanges group large non-retail participants into three buckets, collectively called the san da fa ren (三大法人, literally the three big juridical persons — in English, the three institutional investors):
- Foreign investors (waizi, 外資): overseas funds, sovereign wealth, and global asset managers trading through qualified foreign investor accounts. They are by far the largest bucket and hold dominant stakes in Taiwan’s blue chips — as of July 2026, foreign ownership stood at 69.33% of TSMC, 62.88% of Delta Electronics, and 55.56% of MediaTek (TWSE data).
- Investment trusts (touxin, 投信): domestic mutual funds and ETF managers. Smaller in absolute size, but their buying is concentrated, persistent, and often tied to fund launches and quarter-end positioning, which makes it disproportionately influential in mid-cap names.
- Dealers (ziyingshang, 自營商): the proprietary trading desks of local securities firms, reported in two sub-lines — trading for their own account and hedging (mostly for warrants and ETF market-making). The hedging line in particular is mechanical flow, which is why experienced readers weight dealer numbers lightly.
Where and when the data appears
The Taiwan Stock Exchange (TWSE) publishes the aggregate net trading values of all three groups shortly after each session’s close, in both Chinese and English, on its website and through its open API. Two companion releases matter:
- T86 — the per-stock file showing each group’s net buy or sell in every listed security. This is where you see which names the money went into, not just how much.
- TPEx equivalents — the over-the-counter market (home to many smaller tech names) publishes the same three-bucket breakdown separately. If you only read TWSE data, you miss the OTC half of trust and foreign activity.
The Taiwan Futures Exchange (TAIFEX) separately publishes institutional positioning in futures and options after each session — indispensable for the cross-check described below.
A worked example: reading one day’s report
Here is the actual TWSE release for the session of July 22, 2026 — the session preceding the second day of that late-July rebound:
| Group | Net trading value | Direction |
|---|---|---|
| Foreign investors | NT$+17.34B | Net buy |
| Investment trusts | NT$+18.90B | Net buy |
| Dealers (prop + hedge) | NT$−2.40B | Net sell |
| Combined | NT$+33.85B | Net buy |
Read it in this order:
- Sign of the foreign line first. NT$+17.34B is a solid net buy. As a rough calibration, single-day foreign flows in the low tens of billions of TWD are meaningful but ordinary; prints many times that size mark genuine regime days.
- Compare trusts to foreigners. Here trusts bought NT$+18.90B — slightly more than foreigners, which is unusual given their smaller asset base. That is exactly the signature you would expect if domestic funds were leading a stabilization effort after a retail washout, and it is why commentary at the time credited domestic mutual funds specifically.
- Discount the dealer line. NT$−2.40B is small and dominated by hedging flow. Unless the dealer number is extreme, treat it as residual.
- Now cross-check derivatives. The same session’s TAIFEX data showed foreign investors holding a net short position of 76,595 contracts in TAIEX futures open interest, with a near-flat day’s net volume (+465 contracts). Cash buying alongside a large standing futures short usually means the buying is partially hedged — a real but qualified vote of confidence, not an all-in bullish stance. The options put/call ratio on open interest stood at 118.24%, i.e., more put than call open interest, consistent with hedged rather than euphoric positioning.
Conclusion from this one day, properly read: institutions genuinely supported the market, trusts led relative to their size, and foreigners bought cash while keeping index hedges on. That is a materially more nuanced picture than the headline combined buy of NT$33.85B suggests.
Streaks matter more than single days
One day of foreign buying is weather; ten consecutive days is climate. Local media track consecutive net-buy and net-sell streaks obsessively, and for good reason:
- Foreign flows are persistent. Global allocators move money in programs over days or weeks, so a streak of the same sign is the clearest evidence of an allocation shift rather than tactical trading.
- Streaks in value beat streaks in direction. Five days of small net buys after twenty days of heavy selling is stabilization, not reversal. Track the cumulative sum over the streak, not just its length.
- Trust streaks are episodic. Because trust buying often clusters around fund launches and seasonal positioning, a trust streak tells you more about which sectors domestic managers are committed to than about overall market direction.
Per-stock data: the T86 file and foreign ownership
The aggregate number tells you the tide; T86 tells you which boats it lifted. Three practical uses:
- Confirm index moves. Because foreign ownership is so concentrated (as of July 2026: TSMC 69.33%, Delta 62.88%, MediaTek 55.56%, UMC 41.03%, Hon Hai 40.18% of shares outstanding), a foreign net-buy day that does not show buying in the top index weights is a divergence worth noting — it means the headline flow went into second-tier names.
- Spot trust concentration. Trusts often press a small set of mid-caps for weeks. Sorting T86 by trust net buys over a rolling window surfaces these campaigns early.
- Separate the short channels. Institutional shorting shows up in securities lending (SBL, jiequan, 借券) balances, published per stock by TWSE — distinct from retail margin shorts. On July 22, 2026, for instance, SBL balances rose in 449 shortable issues and fell in 447, an almost perfectly balanced day, while the market-wide retail margin loan balance (rongzi, 融資) rose NT$6.33B to NT$582.8B. Reading both tells you whether pressure is coming from institutions or leveraged retail.
Why foreign flows dominate the narrative — the Seoul parallel
Taipei shares a market microstructure with Seoul: an export-heavy index dominated by one or two chip giants, foreign ownership of roughly 40% of the market, and a currency that moves with the same flows that move the index. When foreigners buy Taiwan equities they generally must buy TWD, so sustained foreign equity inflows and TWD strength tend to arrive together (USD/TWD stood at 32.27 as of July 23, 2026). This creates the feedback loop familiar to Korea watchers: foreign buying lifts the index and the currency, currency strength improves foreign investors’ unhedged returns, which encourages further buying — and the loop runs equally hard in reverse. It is also why the foreign line, not the combined line, leads every Taipei market wrap, exactly as foreign flow leads coverage of the KOSPI.
Common mistakes to avoid
- Reading one day as a signal. Programs span weeks. Wait for the streak.
- Ignoring the futures book. As the worked example showed, cash buying with a large futures short is hedged exposure. Always pair the TWSE cash report with TAIFEX open interest.
- Treating dealer flow as conviction. Much of it is warrant and ETF hedging.
- Forgetting the TPEx market. Trust activity in OTC-listed small caps is invisible in TWSE-only data.
- Confusing SBL shorts with margin shorts. The institutional and retail short channels are separate datasets with different meanings.
- Using the TSM ADR gap naively. TSMC’s US-listed ADR habitually trades at a structural premium to the Taipei shares — roughly 15–25% in recent years, owing to limited fungibility — and the two closes come from different sessions. The change in that gap carries information about foreign sentiment; its level does not.
FAQ
Are foreign investors always the biggest of the three?
By assets and typical daily turnover, yes. But on individual days trusts can out-buy them — as on July 22, 2026 (trusts NT$+18.90B vs foreigners NT$+17.34B) — and those days are worth flagging precisely because they are unusual.
Does foreign net buying predict the next day’s market?
Not reliably on a one-day horizon. Its value is confirmatory and cumulative: sustained streaks coincide with allocation shifts, and divergences between cash flows and futures positioning warn you when a headline number overstates conviction.
Where can I get this data programmatically?
TWSE operates an open API (openapi.twse.com.tw) covering the daily institutional aggregates, T86 per-stock data, margin balances, and SBL balances. TAIFEX publishes institutional futures and options positioning on its own site, and TPEx covers the OTC market.
Is the data revised after publication?
The daily figures are settlement-based exchange records and are effectively final as published; there is no meaningful revision cycle to worry about, unlike macroeconomic statistics.
Sources
- TWSE institutional investors, T86, margin and SBL data — openapi.twse.com.tw and www.twse.com.tw/en
- TAIFEX futures and options positioning — www.taifex.com.tw
- TPEx (OTC market) institutional data — www.tpex.org.tw
- Company filings — MOPS, emops.twse.com.tw
- TSMC investor relations — investor.tsmc.com
Disclaimer: This is an information and analysis publication, not investment advice. See our Methodology for data sources, standards, and our corrections policy.