KOSPI Breaks 7,000 as US$23.2 Billion Walks Out: A Rebalancing Playbook for MPF’s Asia-Pacific Bet

Track 2 · The Data Trading Desk | 2026-09-23 | minami

On 21 September, South Korea’s KOSPI closed at 7,007.72, reclaiming the 7,000 mark; Taiwan’s TAIEX keeps printing record highs. Year to date, the KOSPI is up roughly 61.8% and the TAIEX about 59.3% — the two best-performing equity markets on the planet, powered almost entirely by a single trade: AI semiconductors. But EPFR flow data via Bank of America tells a different story: US equity funds shed US$14.2 billion over the past three weeks, the largest outflow since January, and global equity funds lost US$23.21 billion in the week through 16 September, the biggest weekly outflow in nine months. Prices at record highs, money walking out the door. That divergence is precisely when MPF members holding Asia-Pacific equity funds should be applying rebalancing discipline.

1. The Macro Matrix: A Risk-Off Warning Beneath a Risk-On Surface

Yields — the hiking regime is officially back. The Federal Reserve raised rates 25bp on 16 September, its first hike since 2023, and the HKMA lifted its base rate to 4.25% in step. The US 30-year Treasury yield sits at its highest since June 2007; Japan’s 10-year JGB touched 3.00% for the first time since 1996. Yields are draining equity flows: global bond funds averaged US$17.5 billion of weekly inflows over the past four weeks, while equity funds averaged just US$7 billion a week over the past three — down from US$52 billion a week in July.

Liquidity — smart money is turning defensive. BofA’s private-client data shows 66.2% of assets in equities, 17.2% in bonds and just 9.4% in cash — a record-low cash allocation. That is textbook late-cycle behaviour: everyone fully invested, no ammunition left. Technology sector funds recorded outflows in three of the last four weeks — even as semiconductor earnings came in strong.

Technicals — the index makes highs, the leader blinks first. TSMC’s ADR closed at US$417.72 on 16 September, down 9.4% from its three-month high of US$460.90, despite strengthening fundamentals. That is the first crack inside the AI trade: the index holds its record on the shoulders of a few giants, while the biggest giant has already started to diverge.

Net reading: risk-on on the surface, risk-off underneath — price momentum intact, but liquidity and flows have turned.

2. The Evidence: Three Divergences You Cannot Ignore

Divergence one: US$1.5 trillion spent versus productivity below trend. BofA strategist Michael Hartnett notes that roughly US$1.5 trillion has been poured into AI over the past three years, yet total factor productivity has fallen below trend — a measure highly correlated with consumer confidence over the past 50 years. “Sometimes Main Street knows what Wall Street doesn’t.” When the return on that capex gets questioned, the most richly valued link in the chain corrects first.

Divergence two: the concentration arithmetic. Samsung Electronics and SK Hynix together account for close to 40% of the KOSPI’s market capitalisation (market reports); Samsung alone has crossed a US$1 trillion valuation. In other words, buying an Asia-Pacific equity fund tracking Korea means nearly half your exposure is a bet on two memory-chip makers. South Korea’s August exports surged 68.7% year on year to US$98.25 billion, with AI semiconductor shipments up 209% — spectacular data, and concentrated data: every scrap of good news is already in the price.

Divergence three: MPF winners are concentrated winners. GUM’s Q2 data shows Asian equity funds returned 27.9% in the first half, the best of all MPF categories; a BCT programme revealed members in the same scheme ranging from +32% to −19% over the half — a gap of more than 50 percentage points. The winners’ secret was not stock-picking skill; it was simply standing on the AI-semiconductor track. Being on the right track is luck; still being on it when the track turns is risk.

Translated into Hong Kong dollars: the average MPF balance is about HK$343,420 (MPF Ratings, August). If 30% of that — roughly HK$103,000 — sits in Asia-Pacific equity funds, a 20% sector pullback wipes out HK$20,600, equivalent to four months of HK$5,000 contributions. That is not a forecast; it is the mathematics concentration brings with it.

3. The Switching Blueprint: Three Trimming Plans (Illustrative Framework)

The following is a discipline framework; percentages are illustrative assumptions — adjust to your own risk tolerance before acting:

Plan A — Disciplined trim (for Asia-Pacific allocations of 25% or more)
– Asia-Pacific equity funds: cut 10 percentage points (e.g. 30% → 20%)
– Redirect 5pp into North American equity funds — the genuine diversifier inside MPF, with the lowest correlation to Asian tech
– Redirect the other 5pp into bond funds or MPF Conservative funds as dry powder in a hiking environment

Plan B — Defensive tilt (for those who prefer small moves)
– Keep the Asia-Pacific core, but direct 50% of new contributions over the next three months into Conservative funds
– Let “new money” naturally dilute the concentration, avoiding the transaction cost and unknown-price risk of a lump switch

Plan C — Trend-following trigger (for the systematic)
– Rule: if the KOSPI or TAIEX closes below its 50-day moving average for five consecutive sessions, execute half of Plan A’s trim
– If it reclaims the 50-day average and holds, do not chase back — discipline only subtracts; it never predicts

The common principle across all three: you are trimming excess concentration, not calling the top of AI. AI remains a decade-grade technology wave, but MPF is a 30-year retirement project — it cannot afford a structure with nearly 40% riding on two companies.

4. Unknown-Price Protection: Execution Discipline for T+1/T+2 Blind Trading

MPF fund switching uses forward pricing — the NAV is confirmed T+1 or even T+2, so you never know the dealing price when you place the order. Three iron rules apply:

1. Stage the switch. Split a 10pp adjustment into two or three tranches a week apart, so no single unknown-price day decides the outcome, and to limit out-of-market risk during the transit.
2. Avoid event days. No orders in the two days around Fed decisions or TSMC/Samsung earnings. In the week of the 16 September hike, global equity funds lost US$23.2 billion in a single week — event-day unknown prices are the most expensive.
3. A cooling-off period after each switch. Observe for at least a month; do not chase back on a one-week bounce. BCT’s data on members who switched funds 110 times in half a year is the cautionary tale of paying the unknown-price tax in full.

This week’s action list: ① log in to your MPF account and check whether Asia-Pacific equity funds exceed 25% of your balance; ② check where the KOSPI/TAIEX sit relative to their 50-day averages; ③ if over the line, pick Plan A, B or C and start the first tranche this week. The value of discipline is not in calling the top — it is in not being fully loaded when the top arrives.

Sources: Seoul Economic Daily (2026-09-21), BofA Global Research citing EPFR (mid-Sep 2026), Reuters citing LSEG Lipper (2026-09-18), Yonhap (2026-09-01), GUM (2026 Q2), BCT, MPF Ratings (2026-08). Allocation percentages are an illustrative framework, not investment advice.