China's "National Team" in A-Shares: A Panoramic Study — Reading Its Intraday Footprints and Inferring Its Next Moves

Key takeaways (read this first)

The national team leaves "fingerprints" on the tape, and the 2026 pattern has been validated repeatedly. The core set: a time fingerprint (concentrated in the last 20–30 minutes, now increasingly seen intraday as well), a target fingerprint (CSI 300 > SSE 50 > STAR 50 / ChiNext, with multiple same-family products bought in unison — "group tactics"), a volume gap (5–20× normal), a share-growth fingerprint (buying more as prices fall), and a premium/discount fingerprint (narrow premium = allocation, not speculation).
No single signal is sufficient — you need "two of three." The hardest combination is "tail-end pulse + share growth + concentration in the largest broad-based ETFs." If there is only verbal support with no trace on the tape, it is a "mouth-only bottom" and should not be traded.
In 2026 the playbook shifted in a decade-defining way: from a "firefighter that only acts on down days" to a "market-maker active on both ends." In H1 it systematically reduced broad-based ETF holdings at market highs (net outflow of ¥534.6bn from Jan 15), then stepped back in twice at lows (Jul 13–17 inflow ¥96.1bn). This means "the team is present" ≠ bullish and "the team is reducing" ≠ bearish; its objective function is to damp volatility, not maximize returns.
Tracking got materially harder in 2026. The disclosure rule effective 2026-05-01 abolished the ETF semi-annual top-10 holder list, so Huijin's holdings can no longer be verified from periodic reports; the number of stocks where China Securities Finance (证金) appears in the top-10 holders fell from 77 (26Q1) to 22 (26Q2). Only three substitutes remain: high-frequency ETF share/creation-redemption data, the PBOC's "claims on other financial corporations" monthly line, and Guoxin/Chentong增持 announcements.
The correct inference: it changes the speed of decline, not the trend. The support bid interrupts the "price down → margin contraction → redemption up → price down" feedback loop, but it will not push the index higher on its own. The right use is to calibrate your own position and pace (decelerating downside, converging volatility), not to treat it as a buy signal. The exit is equally gradual: "stop adding → shares stop growing → support bid disappears."

Scope & sources: holdings per Wind's "national team" definition (incl. the social security fund, Central Huijin + Huijin Asset Mgmt, China Securities Finance + its vehicles, and SAFE platforms), via Caijing 2026-09-03 / Sina 2026-09-04; ETF flows from Wind / Choice via China Fund News 2026-10-21, Caixin 2026-07-20. Media-attributed figures are flagged and should be checked against primary sources.

1. Calibration first: exactly "who" is acting

Before reading intraday signals we must fix the scope. At the same point in time, four common definitions can produce answers that differ by 40%. Mixing them leads to the illusion that "the national team could dump ¥4tn at any moment."

Tier	Entities covered	Scale	What it tells you
Narrowest	Central Huijin Investment + Huijin Asset Management ("Huijin complex")	≈ ¥2.85tn	structural, long-horizon market force
Tier 2	+ China Securities Finance (证金) & its vehicles, CSFC bespoke funds	≈ ¥3.3tn	routine counter-cyclical ammunition
Tier 3	+ SAFE platforms (Wutongshu / Fengshan / Kunteng)	≈ ¥3.4tn	long-term FX-reserve allocation
Widest	+ NSSF, basic pension fund, state capital ops platforms (Guoxin / Chentong)	≈ ¥4.6tn	the upper bound of "bullets" society can mobilize in a crash

Source: WeChat "同心成城" "Anatomy of the National Team" 2026-10-02 (media estimate; tier figures are ranges, verify against primary). Wind's ¥3.88tn (26Q2, top-10 circulating-holder basis, 656 names) is a different, non-additive statistic — it omits ETF-channel share purchases entirely.

Six actors, six different tape fingerprints

Actor	Role	Scale & key data	Tape signature
Central Huijin (incl. Huijin AM)	core /主控	2025 year-end total assets ¥9.79tn, net assets ¥8.33tn, revenue ¥973.3bn, net profit ¥940.8bn, gearing 15.0%; paid-in capital ¥828.2bn; 26 controlled/participating institutions	buys broad-based ETFs (CSI 300 / SSE 50 / STAR 50) almost exclusively; barely touches individual stocks (bank stakes are control rights, not trading positions); acts in a few concentrated windows
China Securities Finance (证金)	firefighter /备用接口	2015 rescue lead (21 brokers at 15% of net assets, ≥¥120bn into blue-chip ETFs); top-10 stock appearances: 77 (26Q1) → 22 (26Q2) (55 exited)	being redefined as a standby channel; 26Q2 cut Sinopec 1.552bn sh (+78.46%)
National Social Security Fund (社保)	balanced long money	2025 year-end assets ¥3.81tn, 2025 investment gain ¥390.7bn, return 13.22%, 7.62% since inception; A-share holdings 26Q2: 620 names, ¥536.2bn (Choice)	the only member with fully public returns; most balanced sector mix (hardware equipment leads); adds in electronics/manufacturing/resources
SAFE platforms Wutongshu/Fengshan/Kunteng	FX-reserve onshore vehicle	Wutongshu holds 89.26% of CDB Bank, 27.19% of CDB, 65% of Silk Road Fund; A-share holdings ≈¥40–50bn (old basis), ~88% banks	primary mandate is capital injection into policy banks, not a secondary-market stabiliser; holdings static for years
Guoxin / Chentong	fastest-acting industrial platform	2026-07-19 announcement: Guoxin >¥50bn via repo/relending + matching funds, Chentong ≈¥10bn (≈¥60bn combined); 2025-04 first tranche ¥80bn / ¥100bn	the only actor that regularly discloses exact amounts; buys central-SOEs, tech, and ETFs; announcements themselves are the signal
National Big Fund (大基金)	industrial capital (edge)	≈¥330bn (26Q2), top-3 holdings all semiconductors	industry-driven, not an index stabiliser

Sources: Lianhe Credit "Central Huijin Investment Co. 2026 Rating Report" 2026-07-17 (Huijin financials, primary); NSSF Council 2025 Annual Report 2026-09-01; Wind / Choice; Guoxin & Chentong announcements 2026-07-19.

Why "90% of Huijin sits in banks" matters enormously: Huijin's stakes in ICBC, ABC, BOC and CCB are overwhelmingly control rights of state financial capital accounted as investments, not trading positions. At BOC, Huijin Investment holds 188.792bn shares (58.59%) plus the Ministry of Finance's 8.64% — two-thirds of the bank. Selling even 1 percentage point is hundreds of billions of sell pressure and touches control — so this tranche is technically untradeable. Corollary: to read the national team, do not watch whether banks are sold (they won't be) — watch the ETF channel.

2. ⭐ Intraday reading: the seven "fingerprints"

This chapter is the heart of the report. Signals are ordered by descending reliability, each with: data source, trigger threshold, and how to falsify it. The two tail-end volume explosions of Oct 15 and Oct 20, 2026 are the cleanest empirical samples in years, and we use them as the yardstick — with one crucial 2026 update: the absorption window has shifted forward.

Figure 1 · Empirical tail-volume in the largest broad-based ETFs, 2026-10-20 (¥100m)

Sources: Wind / SSE public data, via National Business Daily, China Fund News, Tencent Finance (2026-10-21). "Last-20-min turnover" is a media-disclosed figure; different outlets differ slightly (e.g. ChinaAMC STAR 50 full-day ¥89.37bn vs ¥97bn) — lower values taken here, non-primary source, verify.

Time fingerprint: not the open — after 13:30, especially the last 20–30 minutes (now also intraday)

high reliability

Pattern

The burst of volume does not come in the morning panic; it starts after the panic has largely exhausted itself, ramps through the afternoon, and explodes in the final 20–30 minutes before close. Increasingly in late October 2026, a comparable "steady absorption" is also visible intraday between 10:00 and 14:30 (see new Fingerprint 1b).

Evidence

2026-10-15: ChinaAMC CSI 300 ETF full-day ¥19.41bn (≈20× prior day), last 30 min >¥12bn; Huatai-PineBridge CSI 300 (510300) full-day ¥16.09bn, last 20 min >¥12.5bn, ≈78% of the day, vs ¥0.41bn in the same window the day before — over 30×. 2026-10-20: ChinaAMC SSE 50 (510050) full-day ¥7.53bn (+1.98×, a 9-month high since 2026-01-29), last 20 min >¥4.29bn, ≈80% of the day.

Why it works

Retail, quant, mutual funds and northbound money never pile buy orders into the last 20 minutes aimed at the same index basket; the time-target double concentration is the behaviour of an allocation-sized player.

Falsify

If the burst is only in the final 3-minute auction, it may just be institutions minimising impact cost — not a fingerprint. If it is in the morning, it is more likely sentiment or event-driven.

NEW (2026-10) · Intraday absorption: the window has shifted forward

emerging pattern

Observation

Institutional observers note that broad-based ETF absorption has shifted from the tail to intraday (reported 2026-10-25). Watching only the 14:30 close would systematically miss this round of support. The footprint: in the 10:00–14:30 window, the minute-average turnover runs ≥1.8× the day's median, with cumulative turnover ≥¥500m and the window persisting ≥30 minutes.

Why it matters

The "product rotation, no announcement, absorb on the way down" Huijin style means the same buyer can show up at 10:30 as well as 14:50. Treating 14:30 as the only tell undercounts the real presence.

Caveat

This is a newer, not-yet-consensus read. The dominant visible fingerprint is still the tail burst; intraday absorption should be used as a supplementary confirmation, not a standalone trigger.

Target fingerprint: broad-based first, and same-family products in "group tactics"

high reliability

Priority

CSI 300 > SSE 50 > STAR 50 / ChiNext > CSI 500 / CSI 1000 (occasional, weaker). On 2026-10-20 by index: CSI 300 net inflow ¥6.889bn led, SSE 50 ¥4.44bn — the two core broad-based ETFs supplied most of the increment.

Group tactics

Multiple ETFs tracking the same index surge together: on 10-20, ChinaAMC SSE 50 (+¥4.419bn), ChinaAMC CSI 300 (+¥3.982bn), Harvest CSI 300 (+¥1.571bn) and Huatai-PineBridge CSI 300 (+¥0.865bn) all made the leaderboard — a "group" no single institution or arbitrageur can manufacture.

Growth tilt (2026 update)

This round's buying has tilted toward growth: per Bloomberg Intelligence (2026-10-22), STAR 50 + ChiNext now make up ≈32% of national-team buys (vs 11% historically), while CSI 300 has fallen to ≈45% (vs 55%). The real net-inflow ranking of this round (Sep 28–Oct 23) confirms it: Huatai-PineBridge CSI 300 ¥47.5bn > ChinaAMC STAR 50 ¥39.8bn > E-Funda ChiNext ¥38.5bn > E-Funda CSI 300 ¥26.9bn; SSE 50, Harvest/ChinaAMC CSI 300, Southern CSI 500 and Southern CSI 1000 each >¥12bn. STAR 50 and ChiNext are now main battlefields, not sidekicks.

Falsify

A single product spiking with no movement in its peers = idiosyncratic subscription or market-maker behaviour; a spike concentrated in sector/themed ETFs = industrial or thematic money, not stabilisation money.

Share fingerprint: buys more as prices fall — real primary-market creation (the hardest evidence)

highest reliability

Mechanism

Secondary-market ETF trading only churns existing shares — total shares are unchanged. Only primary-market creation lifts total shares. "Share growth" is the only hard line separating "someone is buying" from "real new money is arriving."

Evidence

Week of Oct 12–16: all ten ETFs with the largest share gains fell — ChinaAMC STAR 50 −7.2% but shares +7.218bn; HuaAn ChiNext 50 −9.32% but shares +1.422bn. "Buy regardless of direction" is the stabiliser's signature, distinct from elastic bottom-fishing capital.

Continuity

Sep 28–Oct 23: stock ETFs saw 15 straight sessions of net inflows totalling ¥319.8bn, of which ¥295.5bn (92%) went to broad-based; on Oct 22 alone broad-based net inflow was ¥51.6bn.

Falsify

Shares flat but turnover soaring = churn, no new money; share growth concentrated in small or themed products = ordinary rebalancing.

Premium/discount fingerprint: narrow premium = allocation, wide = spec

mid-high reliability

Mechanism

True long money cares about NAV-level allocation and will not push the secondary price to a visible premium just to get filled; short-term hot money creates wide premiums that attract arbitrageurs whose subsequent selling adds pressure.

Read

In the Oct 2026 tail bursts, the leading CSI 300 ETFs held premiums in the low-per-mille range — dominant money was allocation, not speculation.

Extend

In the 2015 and 2025-04 extremes, ETF prices above IOPV pulled arbitrageurs to buy stocks and swap for ETFs — the classic "indirectly create stock volume without hitting individual names" mechanism. Persistent significant premium + simultaneous component volume often signals an escalation of support.

Structural divergence: index vs stocks, vs margin, vs northbound

mid reliability

Yellow-white line

Index turns green or narrows its loss while most stocks stay red = only weights are being supported. On 2026-10-15, >5,100 stocks fell yet turnover expanded to ¥1.63tn (+¥218.6bn) and the index reclaimed 3,700 into the close.

Margin divergence

Margin balance falling while the index holds = the "external force" (not retail leverage) is catching the falling knife. By 2026-10-26, margin balances had fallen from ¥2.65tn (early Sep) to ≈¥2.46tn, with 11 straight sessions of net selling — leverage retreat running opposite to passive-fund absorption.

Northbound divergence

When tail-end ETF sweeps happen while northbound is still a slight net seller, the two point different ways — proving the buyer's logic is independent of overseas FX and US overnight moves: independent domestic long money.

Announcement fingerprint: Huijin / Guoxin / Chentong statements + PBOC balance sheet (low-frequency, highest certainty)

high reliability (lagged)

Huijin statement

Template phrasing: "firmly bullish on China's capital-market prospects, fully recognises current A-share allocation value, has again increased ETF holdings, and will continue to do so." Historical triggers: 2023-10 (four big banks + ETFs), 2024-02-06 (broadened ETF scope), 2025-04-07/08 (intraday statement, first explicit "quasi-stabilisation-fund" wording). 2026-10-23: Huijin issued an official statement that it bought ETFs that day and will continue to increase holdings — the first explicit official confirmation of this round of action (the Oct 15/20 bursts themselves had none).

Guoxin / Chentong

The only actors that regularly disclose exact amounts: 2025-04 first tranche ¥80bn / ¥100bn; 2026-07-19 Guoxin >¥50bn + Chentong ≈¥10bn. Frequency and size directly track the SOE-platform weight in the stabilisation system.

PBOC line

"Claims on other financial corporations" records PBOC funding to non-bank institutions — the only high-frequency, public, strongly correlated proxy. ≈¥130bn (2023-12) → ≈¥300bn (2024-02) → ≈¥740bn (2024-09) → ≈¥1.03tn (2025-04, +¥370bn in a month) → ≈¥800bn (2025 year-end). A fast rise ≈ emergency entry; a fall ≈ exit or self-healing.

Single-stock fingerprint: the four big banks "won't fall" + "证金-held" concept stocks (significantly decayed in 2026)

low reliability (signal degraded)

History

The four big banks not falling was the traditional Huijin entry; in 2015 the team bought individual stocks across >half of listed companies.

2026 reality

证金's top-10 appearances fell from 77 (26Q1) to 22 (26Q2); core anchors (BOC, CCB, China State Construction, PetroChina) retained, the rest withdrawn. The paradigm shifted from "buy stocks" to "support the index" — a broad-based ETF basket covers the whole market and avoids any holding change being read as a value endorsement of a specific company ("save the market, not a stock").

Usage

Stop hunting the national team in top-10 stock lists; this线索's information content has collapsed in 2026. "证金-held concept" still tags 256 stocks (2026-05-08) but that is a label, not a current holding.

One-line rule: an announcement ≈ real; tail pulse + share growth + concentration in top broad-based ETFs ≈ very likely real; only verbal support with no tape trace ≈ a mouth-only bottom.
Mind the boundary: turnover data can prove "a big buyer is buying" but cannot prove who the buyer is. Even after the 2026-10-23 Huijin statement, most days' bursts carry no official confirmation — the read is a highly-consistent inference, not fact.

3. From signal to stance: a six-state regime model

Once fingerprints are spotted, the key is translating "it is buying today" into "what it will do next." The national team's objective is to damp volatility, not maximise returns — so its state machine has six states, and transitions are gradual.

State	Observable features (≥2 required)	Implication & inference	Historical window
A · Unilateral support	① tail (after 14:30) burst in top broad-based ETFs, ≥5× normal ② ETF shares keep growing while the target still falls ③ CSI 300 / SSE 50 net inflow leads, multiple same-family products in "group tactics" ④ narrow premium in per-mille range ⑤ scope expands from large-caps to STAR/ChiNext	Speed bump, not engine. Interrupts the negative feedback, slows the fall, converges volatility — but will not push the index up. Inference: down-slope flattens, "fall – tail reclaim" patterns repeat; not a trend reversal. Watch next morning for "tail lift – next-day chop" tug-of-war between short-term and allocation money.	2024-02, 2025-04, 2026-07, 2026-09/10
B · Oversold-rebound support	① index rising but still >3% below its 30-day high ② broad-based volume elevated on the up-day ③ rebound driven by absorption, not a fundamentals flip	A rally from deep lows with elevated broad-based volume is absorption-style rebound, not high-level distribution. Still needs a 2nd/3rd fingerprint to confirm. Do not read it as "the bottom is in."	2026-10-23 rebound (+1.71% on CSI 300 after 10-week decline)
C · Silent / fading	① tail pulse gone, broad-based ETF volume back to normal ② shares stop growing but don't shrink (subscriptions pause) ③ net inflow turns zero then slightly negative ④ no new announcements	Support bid gone ≠ dumping. The exit is "stop adding → slowly sell," the impact is a bid gap, not active selling. Most misread as "bad news is priced" — actually the floor beneath thins and the index grows more sensitive to fundamental negatives.	2026-08 (media: stock ETF net outflow ¥119.9bn)
D · Counter-cyclical distribution suspect	① broad-based ETF sustained net outflow and shares keep shrinking ② high-weight sectors (banks) clearly lag, valuation repair blocked ③ 证金 top-10 scale contracts ④ happens at market highs / overheated sentiment	A new state that only appeared in 2026, and the easiest to misread. It is not bearish — it is "suppress overheating, guide a slow bull." Inference: up-slope capped, weights lag growth; once the reduction ends, the repressed high-weight sector (banks) gets a valuation-repair window.	2026-01 ~ 06 (net outflow ¥534.6bn from Jan 15)

Position is read first, then volume is interpreted. The national team is state-dependent: it buys when the fall is fast, and stops when the fall slows or the market goes sideways (Zhu Zhenxin, 2026-10). So every assessment first locates "where are we": CSI 300 day-K context — trend over 3/5/10/20 days, drawdown from the 30-day high (weak if ≥3% below, deep if ≥8% below), and whether the last 3 days have rebounded ≥2% (if so, the buy bid's probability of fading rises). A burst at a high level = likely distribution; a burst after a deep decline = likely support. Never read volume in isolation from position.

Figure 2 · 2026 national-team "breathing": from high-level distribution to low-level replenishment (¥100m)

Period	Direction	Scale (¥100m)	Trigger	Source
2026-01-15 ~ 06	net outflow	5,346	market highs (year high 4,197; media says mid-May 4,258), turnover near ¥4tn, regulatory cooling	Wind / Gelonghui (est.)
2026 H1 (ETF est.)	reduce	≈15,000	semi-annual reports stop disclosing specific holders	Sina "2026 H1 specific-institution behaviour"
2026-07-13 ~ 17	net inflow	961.14	global AI divergence + US-Iran geopolitics + US yields up; A-share tech crowded trade retraced	Wind / Caixin 2026-07-20
2026-07-19 ~ 20	announced add	≈600	Guoxin >¥50bn (relending) + Chentong ≈¥10bn	both companies' announcements
2026-09-28 ~ 10-23	net inflow	3,198.08	index down ≈14% from May high; ChiNext 4-month max drawdown ≈38%	Wind / China Fund News 2026-10-21

Read: 2026 is the first time the national team completed a full "high spit – low absorb" loop. This proves two-way capability — and means "seeing it buy" no longer equals "it thinks here is the bottom"; it only equals "it thinks the fall is too fast."

The key constraint: it fixes "too fast," not "should rise"

A sideways comparison makes it clear. On 2026-10-15, the Nikkei 225 fell 4.90% and Korea's KOSPI fell 7.57% (intraday circuit breaker), SK Hynix −11.06%, Samsung −9.48%; the Shanghai Composite fell only 1.59% to 3,707.48 — far less than Japan/Korea, thanks to intraday absorption. But A-shares falling less does not mean A-shares are stronger.

So 3,700 is accurately described as: an "absorption zone" the national team recognises — absorbing the speed of decline, not the trend of decline. Calling it a policy bottom is mistaking inference for conclusion.

4. Holdings panorama: where the money sits (26Q2 map)

Figure 3 · National-team A-share holdings by quarter (¥tn)

Source: Wind. 25Q4 ¥4.88tn; 26Q1 ¥4.41tn / 57.46bn sh; 26Q2 ¥3.88tn / 55.53bn sh. Different bases are not cross-comparable; read within the same basis only.

Figure 4 · 26Q2 sector structure (share)

Source: Wind / Caijing 2026-09-03, Sina 2026-09-04. State big-banks ¥3.2tn, brokers ≈¥0.20tn, insurers ≈¥0.16tn; financials ¥3.56tn (91.8%); non-financial only ≈¥0.32tn.

Top-10 holdings by actor (26Q2, top-10 circulating-holder basis)

Actor	Top-10 / key holdings	Quarter action
Central Huijin Inv.	BOC ≈¥1.14tn (188.792bn sh) · ABC ≈¥1.01tn (140.5bn sh) · ICBC ≈¥0.99tn · CCB · BOCOM · CICC 69.99bn · CSC 69.36bn · New China Life 58.32bn · Ping An 53.04bn · PICC 41.31bn	core holdings essentially frozen; non-financial at lowest since 2015
China Securities Finance	CSC 69.36bn · Ping An 53.04bn · GD Power 4.19bn · China Coal 2.49bn · Nanshan Al 2.35bn · Ningbo Bank 2.24bn · Oriental Yuhong · THS · Luzhou · Midea (142m sh / ¥10.7bn)	cut Sinopec 1.552bn sh (−78.46%); added materials, tech hardware, capital goods
Social Security	ABC 142.77bn · ICBC 87.19bn · PICC 38.10bn · BOCOM 19.97bn · Founder Sec 8.36bn · China Jushi 7.52bn · Guosen Sec 4.56bn · CMPort 4.14bn · BYD 4.12bn · Wanhua 3.73bn	new 179, add 175, cut 169, exit 170; new Zijin 81m sh, Sanan 1.56bn, Goke 1.46bn, JSG 1.18bn
Huijin AM	infra/consumption non-financial: China Railway Grp (230m sh, 4th holder) · CRCC (139m sh) · Shuanghui (30.9m sh / 0.89%) etc.; ≈¥15.11bn, 7 stocks in top-10	holdings unchanged
Guoxin Holdings	China Uranium 12.31bn · Naura 10.21bn · Zijin 8.31bn · China Energy Eng · SMIC · Kingsemi · Riyue · NARI (new) · China State Construction (new) · CNNC (new)	clearly tilted to electronics, comms, tech manufacturing
Basic Pension Fund	Health 100 105m sh · COSL 85.8m sh · Zhejiang Longsheng 78.5m sh · CRRC 72.2m sh (new) · GF Sec 64.8m sh (new); ≈¥43.70bn	growth-tilted, active offensive rebalancing

Source: Wind / East Money Choice via Caijing 2026-09-03, Sina 2026-09-04, Toutiao finance 2026-09, CSN 2026-08. Non-primary, verify. Note: a wider "Central Huijin (incl. AM) ¥4.48tn" basis exists and is not comparable to the ¥3.33tn top-10 basis.

Marginal change: money didn't move, direction did

Top-5 sectors by shares added: optical & optoelectronics (+136m sh, #1), chemical products, professional engineering, IT services, medical services (all >34m sh); the top-10 also include automation, general equipment, semiconductors — tech & "AI+" take six seats.
Sell side: Sinopec −1.552bn sh (−78.46%, leaving 426m), China Coal −81.6m sh (−28.55%), Ping An −107m sh, Nanshan Al (SSF 114 combo −21.29%), China Jushi (two SSF combos −26m sh). Logic: high oil but weak product demand, refining margins squeezed.
Scale caveat (important): after the ¥3.56tn financial block, non-financial is only ≈¥0.32tn; tech additions are mostly tens-to-hundreds of millions of shares — they barely move the weight structure. Their meaning is in direction and continuity, not size. Do not conclude "the national team is loading up on tech."
Baijiu has fallen out of the top-10 sectors on weak tape plus sustained reduction.

5. Behaviour history: six interventions, wildly different outcomes

Round	Trigger	The action	Policy bottom → market bottom	Reusable lesson
2008	global financial crisis, SHCOMP 6,124 → 1,800	Huijin 9/18 announces open-market add of ICBC/CCB/BOC; cuts to RRR & rates, stamp-tax halved	surge next day, then relapsed; 10-28 bottom at 1,664, ~40 days later	Huijin entry = policy bottom flag, not market bottom
2015	off-balance-sheet leverage blown up, thousand-limit-down, 5,174 → 3,373	证金 + Huijin + 21 brokers (15% of net assets, ≥¥120bn into blue-chip ETFs); bought individual stocks across >half of listed companies	policy bottom ~7/9, rebound to 4,184 in a month; 2016-01 circuit-breaker bottom at 2,638, ~half a year later	trillions can stop a stampede, but not substitute for leverage clearing
2018	US-China trade + pledge blow-ups, SHCOMP 2,449	not blanket buying, precision relief: local SOEs, Chentong, Guoxin cut the pledge-risk chain	10-19 +2.58%, rebound above 2,700 in a month; 2019-01-04 bottom at 2,440, ~2.5 months, no new low	flood irrigation worse than precision drip; opened 2019–2021 structural bull
2023-10	property down + external hikes, CSI 300 low valuation	Huijin first adds the four big banks (≈¥476m combined), then announces ETF buys and continued increases	gentle, banks short-term lifted, but fundamentals unrelieved, index chopped	stabilisation tools extend from stocks to index products (starting point)
2024	persistent grind, SHCOMP below 2,700	Feb Huijin broadens ETF scope/force/scale; broad-based ETFs become the main entry; >¥410bn net ETF inflow, ~¥876bn added for the year	ETF support directly improves index-level liquidity supply-demand, A-shares rebound fast	broad-based ETFs become the key entry; Sep-24 swap-facility + repo-relending institutionalises stabilisation
2025-04	US reciprocal tariffs, 4/7 SHCOMP −7.34% (3,096.58)	Huijin intraday statement adds ETFs, first explicit "quasi-stabilisation-fund" framing; PBOC backs with relending; Guoxin ¥80bn / Chentong ¥100bn; 48 central SOEs announce buybacks	4/8 SHCOMP +1.58%, ChiNext +1.83%; sentiment repaired far faster than 2015	policy-communication cost dropped; counter-cyclical action becomes standard
2026	H1 overheated highs → Jul external shock → Sep/Oct sustained adjustment	H1 systematic high-level reduction (net outflow ¥534.6bn from Jan 15); two days of ¥100bn+ adds in Apr volatility (media); Jul 13–17 inflow ¥96.1bn + Guoxin/Chentong ≈¥60bn; Sep–Oct renewed sustained inflow	index down ≈14% from May high; Oct finds absorption near 3,700; 10-23 official Huijin statement	from "firefighter" to "market-maker": acts on both ends

Sources: CCTV Finance 2026-07-21, China.com 2026-07-21, People's Daily "What is a stabilisation fund" 2025-04-30, Securities Times 2024-02, Xinhua 2025-04-08, China Industry News 2026-07-30.

Figure 5 · PBOC "claims on other financial corporations": an institutional thermometer of stabilisation money (¥100m)

Source: market researchers' compilation from the PBOC balance sheet, via WeChat "同心成城" 2026-10-02. This is inference, not official confirmation — the PBOC has not publicly linked this line to Huijin operations; use as a directional proxy only. 2025 year-end ≈¥800bn is an estimate.

6. Three (plus one) structural shifts in 2026 that force a new tracking method

Shift 1: from "buy only on down days" to "act on both ends"

−5,346亿

broad-based ETF net outflow from 2026-01-15

CSI 300 −2,673 / CSI 1000 −715.77 / SSE 50 −491.09 / STAR 50 −456.72; sector ETFs +1,262 contrarian

−77.5%

Huatai-PineBridge CSI 300 (510300) size drop

¥422.26bn (start) → ¥94.87bn (end-Jun); Huijin shares 73.513bn (82.76%) → 18.915bn

−1.28tn

whole-market ETF size drop in H1

¥6.02tn → ¥4.74tn (−20%+); product count rose to 1,595 (+193)

DX Securities' read is worth quoting: in Jan–Feb and Apr–May 2026, with sentiment hot and some sub-sectors overheated, the national team reduced major broad-based ETFs to guide a "slow bull." That is the opposite direction from "support" — the former prevents falls, the latter suppresses rises. CICC's addition is equally key: national-team "fade" ≠ "exit"; its stabiliser function is unchanged.

Shift 2: from "buy stocks" to "support the index"

证金's top-10 appearances fell from 77 (26Q1) to 22 (26Q2). A broad-based ETF basket covers the whole market, ending single-point weight support and its market divergence; more importantly it avoids any holding change being read as a value endorsement of a specific company — "save the market, not a stock," killing the selective-intervention controversy. The retreat is incomplete: BOC, CCB, China State Construction and PetroChina anchors remain.

Shift 3: the tracking window was narrowed by regulation

On 2026-03-13 the CSRC released a revised disclosure format rule (No. 2) effective 2026-05-01, abolishing the ETF periodic-report top-10 holder list. Previously investors could track how many shares Huijin held in a given CSI 300 ETF — at 2025 year-end Huijin Investment + Huijin AM held 73.513bn shares (82.76%) of Huatai-PineBridge CSI 300, the single most reliable public clue to national-team posture. From the 2026 interim reports, that clue was shut off institutionally.

The legislative intent was to implement the "high-quality mutual-fund development" plan and follow overseas practice, breaking "copy-trade thinking"; the objective effect is the largest holder's behaviour vanished from public view. The impact on tracking is fundamental: only three substitute clues remain — high-frequency share/creation-redemption data + tail-volume structure + announcement events.

Shift 4 (2026-10): the absorption window moved forward

Beyond the regulatory narrowing, the timing of absorption itself shifted. Where 2024–early-2026 support was a near-pure tail-end phenomenon, late-October 2026 observers report broad-based ETF absorption now also appears intraday (10:00–14:30), with product rotation and no announcement. The monitoring model therefore adds an intraday-absorption fingerprint (steady ≥1.8× median volume over ≥30 minutes, ≥¥500m cumulative) and treats "group intraday absorption across ≥3 broad-based ETFs" as equivalent evidence to a tail pulse. Watching only 14:30 would undercount this round.

7. Three most common misconceptions

Myth 1: "The national team is a stabilisation fund"

Huijin was founded 2003-12 with the statutory duty to "make equity investments in key state financial enterprises under State Council authorisation," explicitly "not conducting any other commercial activity." It only called itself a "quasi-" stabilisation fund's "function" in April 2025 — note the two qualifiers. The difference is not wordplay: a holding company's balance sheet is control rights (mostly untradeable); a stabilisation fund's is tradeable securities (in and out at will). Conflating them overstates usable firepower.

Myth 2: "The national team defends every day"

Reality: on the vast majority of sessions it does nothing. Holdings were flat all 2025; 2026 action concentrated in the Jan–Feb and Apr–May reductions, the July add, and the Sep–Oct add. Daily A-share trading is dominated by quant (≈30% of turnover), retail (≈53%) and mutual funds. Its influence is not in volume but in directional signal and absorption capacity at the crucial moment. Explaining everyday tape by "is the team defending today" is textbook reverse causation.

Myth 3: "Follow the team and you'll make money"

Three non-replicabilities: information lag (stock holdings quarterly, ETF holdings undisclosed since 2026 interim — the window is usually closed when you see it); different objective function (it can keep adding at a loss, no redemption/ranking/stop-loss pressure); different capital cost (it can access PBOC relending, policy-priced). The bigger mismatch is on the sell: you sell at your profit target, it sells because the market is too hot. National-team reduction often happens exactly when sentiment is most bullish.

Three real costs that must be counted

Surrender of price discovery. When ~80% of a broad-based ETF sits with one holder, marginal index pricing power contracts from thousands of independent judgements to one policy-objective decision process. This mechanism is permanent — it does not vanish when markets recover.
Moral-hazard accumulation. Markets repeatedly verify "fall to a level and money catches it," so investors price that implicit guarantee into their risk models and systematically overestimate resilience in extremes. The H1 2026 high-level reduction was partly to suppress exactly this expectation — support must exist but not be taken for granted (Lian Ping, China Chief Economist Forum, Jul 2026).
Exit-transparency gap. The new rule lets ¥4tn-scale money churn positions unobserved, institutionally amplifying information asymmetry. This is the most worrying item today.

8. Actionable tracking checklist

Daily (within 15 min of close)

Indicator	Trigger
Top broad-based ETFs (510300/510310/159919/510050/588000/159915) last 20–30 min share of turnover	>50% and >3× vs prior day
Multiple same-index ETFs surging together	≥3 in same-day net-inflow top 20
CSI 300 / SSE 50 ETF share change	shares up while target down
Intraday absorption (10:00–14:30) in top broad-based	≥1.8× median vol, ≥¥500m, ≥30 min
Yellow-white line divergence, margin direction	index narrows loss while stocks fall, margin drops

Weekly / monthly

Indicator	What to watch
Broad-based ETF inflow/outflow continuity	≥5 consecutive same-direction days = allocation; a single day is noise
Guoxin / Chentong add statements & relending use	frequency and per-event size
PBOC "claims on other financial corporations" monthly	fast rise ≈ emergency entry; fall ≈ exit or self-heal
证金 top-10 stock count (quarterly)	stable near 22 = paradigm shift complete
Huijin statement / CSRC stance / forums	"studying more incremental policy tools" wording

Current read (2026-10-26): CSI 300 posted a record ten consecutive weekly declines (longest since its 2005 inception; the modern record was six weeks in Dec 2023). On 10-23 it rebounded +1.71% to 4,105.33, ChiNext +4.16%, two-market turnover expanded to ¥1.97tn. Margin balances have fallen from ¥2.65tn (early Sep) to ≈¥2.46tn with 11 straight net-selling sessions — leverage retreat runs opposite to passive-fund absorption. Bloomberg Intelligence estimates Huijin raised ≈$231bn via ETFs in H1, redeployed only ≈$43bn since July, leaving ≈$188bn dry powder; Oct 9–20 inflow was ≈$15bn, slower than July — batch dip-buying, not lifting. On 10-23 Huijin issued an official statement confirming the ETF purchase.
Around this window: the 6th Plenum (Oct 26–29), the Fed decision (Oct 29, ~90% hold, ~67.5% hike by Dec) and the SSE index-revision (single-sample cap 15%, effective Oct 26). By the 10-26 intraday read, CSI 300 was ≈−11.2% from its 30-day high and −1.17% over 3 days with 72.9% of stocks down and no volume fingerprint → the model reads SILENT; the recent rebound also implies the buy bid may be pausing (rebounding → fade).

The counter-signal (must hold both in mind): this round's inflow may also mix in "allocation money systematically contrarian-buying as valuations entered an attractive zone" — not necessarily stabilisation money, but insurers, pensions, wealth-management subs and other medium-long money adding naturally. CICC's Yao Ziwei calls it "consistent with large institutions' equity allocation style," using "consistent with," not "confirmed." UBS's Meng Lei warns: since 2024, whenever the SHCOMP quickly retraces below a key level, ETFs get sustained net inflows — meaning the signal's information content is being front-run and decaying. If you enter only after everyone is watching the tail burst, the excess return is most likely already arbitraged away.

Disclaimer: The above is based on public data and quantitative analysis, for reference only, and does not constitute investment advice. Markets involve risk; invest with caution. Any investment decision should be made independently based on your risk tolerance, financial situation and objectives, and professional advice sought where necessary. Past performance is not indicative of future results.

Sources & data gaps: Primary — National Social Security Fund Council 2025 Annual Report (2026-09-01); Lianhe Credit Huijin 2026 Rating Report (2026-07-17); PBOC, CSRC, Huijin, Guoxin, Chentong public statements; SSE/SZSE (index & market cap, primary). Media (non-primary, flagged) — Caijing, Caixin, China Fund News, Securities Times, Xinhua, People's Daily, CCTV Finance, NetEase Finance, Bloomberg (BI), stheadline. Gaps: ① 2026 Q3 reports still disclosing; no comparable full national-team Q3 holdings; ② ETF top-10 holders undisclosed since 2026 interim; H2 ETF action only inferable from share/creation data; ③ Wutongshu-series no new 2026 periodic report; ④ figures marked "media estimate" are non-primary.