Do US rates sink Hong Kong stocks, and does Beijing rescue them?
Hong Kong's main stock index stands 13% below its January high, as the US 10-year yield climbed to its highest since 2002 and oil passed $110. We tested two common readings of its moves, that US rates drag it down and that Beijing rescues it, on the Hang Seng since 1990.
Do high US rates sink the Hang Seng, and does Beijing rescue it?
On October 6, 2026 the Hang Seng Index closed at 24,281, 13% below its January high (Hang Seng Indexes, 2026b). Over the year the yield on the 10-year US Treasury note climbed from 4.18% to 5.27%, its highest since May 2002 (US Treasury, 2026b). When the index fell 2.6% on October 2, Hong Kong’s public broadcaster put it down to rising oil prices and bond yields (RTHK, 2026). Two days after it fell 13% in April 2025, it steadied as traders bet that Beijing’s state funds would buy (South China Morning Post, 2025). The claim tested here joins those two readings: high US rates, imported through the Hong Kong dollar’s peg, drag the index down, and Beijing steps in when it falls far enough.
A tenth of a percentage point on the 10-year yield, ten basis points, is an ordinary weekly move: since 2022 the yield has moved that much or more in 46% of weeks. A typical six weeks of the Hang Seng, by contrast, moves the index about 5% up or down.
The claim holds only for its first half, and only since 2022. From 1990 to 2021 the Hang Seng rose, on average, in weeks when US yields rose. Since 2022 each tenth of a point on the 10-year yield has come with a 0.35% fall in the same week (Fig. 1). The 3-month Treasury bill, which tracks the Fed’s rate that the peg imports, shows the same pull, so the drag fits the peg and long-term rates alike. The rescue does not show in the index. After the twelve falls into a bear market since 1992, it gained a median 2.0% over the next six months, against 4.7% for any six months. Beijing announced some form of support before the low in each of the ten that kept falling, usually months before it.
Until 2021 the Hang Seng rose in weeks when US yields rose; since 2022 it has fallen.
The Hang Seng's average move in a week for each 0.1-point rise in the US 10-year Treasury yield that week, percent, over all weeks from 1990 to October 6, 2026 and in three eras; bars right of the line are rises.
How we tested it
The test, its definitions and its verdict rules were written down and fixed on October 6, 2026, before the data for it was pulled. The claim has two halves, and each was given a bar to clear (Fig. 2).
The first half is about rates. For every week since January 1990 we took the Hang Seng’s change from one Friday close to the next, and the change in the US 10-year Treasury yield over the same week, in basis points. A regression of the first on the second gives the index’s average move for each 0.1-point rise in the yield. That half holds if the move is negative, and clear of chance at the usual 5% level, both over the whole sample and since 2022. The same test runs on six-week windows, the span from October 6 to November 18, and for three eras: 1990 to 2007, 2008 to 2021, and 2022 on.
The second half is about rescues, so it looks at what the index did after it fell. A sharp fall is a drop of 5% or more over a month of trading, about the size of the drop to October 6. A bear market starts on the first close 20% or more below the highest close of the 252 sessions before. That rule needs a full year of closes, so the fall of September 1990, which came nine months into the series, is not counted, and the first bear market we date is in December 1992. Its low is the lowest close of the following 252 sessions. The half holds if the index, on average, did better than usual in the six weeks after sharp falls, and in the six months after the start of a bear market. For each bear market we also looked up the first measure Beijing announced to support growth or share prices, and when it came against that low.
The rates half holds only since 2022, and the rescue half does not hold.
Each half of the claim, the bar it had to clear, fixed in advance, and the result; the rates test uses every week from January 1990 to October 6, 2026, the fall tests every session from 1990 with a full year or month behind it.
| Half of the claim | Holds if | Result |
|---|---|---|
| High US rates drag the index down | It falls in weeks when the 10-year yield rises, over 1990 to 2026 and since 2022 | Since 2022 only: -0.35% a tenth of a point, against +0.23% over 1990 to 2026 |
| Beijing rescues it after big falls | It does better than usual after 5% monthly falls and after crossing into a bear market | No: 0.86% against 1.09% in six weeks, 2.0% against 4.7% in six months |
Three of the supporting tests written down in advance could not be run as planned. Hong Kong’s exchange publishes daily records of mainland buying through Stock Connect for the past seven months only, so the test of whether mainland money buys the dips is left out (HKEX, 2026b). The HKMA’s interbank rates could not be pulled, so the peg is tested through the 3-month Treasury bill, which moves with the Fed’s rate, rather than through Hong Kong’s own rates. China’s statistics bureau publishes its GDP release dates from 2015, so the GDP test starts there and not in 2010. The bear-market rule, as written, leaves out the fall of September 1990. Several tests were added after the first results came in: the Fed’s decision days, the 2-year yield, the six-week windows split by the size of the yield move, other bear-market lines, and the largest daily moves. Each is reported as an added check.
The index closes are Hang Seng Indexes’ own for the five years it publishes, and Yahoo Finance’s closes for the years before. The two copies match to the cent on 1,227 of the 1,229 days they share, and on the other two days we use the publisher’s figure (Hang Seng Indexes, 2026b; Yahoo Finance, 2026). Yields are the Treasury’s daily par curves, and Brent is the EIA’s daily spot price (US Treasury, 2026a; US Treasury, 2026b; EIA, 2026). The Fed’s decision dates come from its meeting calendars and its record of rate changes (Federal Reserve, 2026a; Federal Reserve, 2026b), and China’s GDP release dates from the National Bureau of Statistics’ yearly release calendars (NBS, 2025).
The Hang Seng is still below its 2007 high
The Hang Seng is a price index of 93 of the largest companies listed in Hong Kong, published since 1969 (Hang Seng Indexes, 2026a). In August 2026 its three biggest members were HSBC, Tencent and Alibaba, and mainland Chinese companies made up at least 65% of its weight (Hang Seng Indexes, 2026a). HSBC and AIA, the largest of its Hong Kong companies, sit beside China’s internet platforms, state banks and carmakers.
The index has had a hard two decades (Fig. 3). It closed at 31,638 on October 30, 2007, and stayed above that level only for three weeks, in January and February 2018, when it set its record of 33,154. It then lost more than half its value, to 14,687 on October 31, 2022, and was back near that low in January 2024. From there it recovered to a January 2026 high of 27,968. On October 6 it stood 23% below its 2007 peak. Since 1992 it has crossed into a bear market twelve times, roughly once every three years.
The Hang Seng has crossed into a bear market twelve times since 1992 and stands 23% below its 2007 high.
The index at each month's last close, January 1990 to October 6, 2026, with a dashed vertical on each day it first closed 20% or more below its highest close of the previous year.
The yield on US 10-year debt has made the opposite journey. It fell from about 8% in 1990 to 0.52% in August 2020, and has climbed since, most steeply in 2022 and again this year (Fig. 4). On October 6 it stood at 5.27%, a level it last reached on May 17, 2002. The Federal Reserve raised its target on September 16, 2026, its first increase since July 2023 (Federal Reserve, 2026b).
The US 10-year yield is back where it was in 2002.
The yield on the 10-year Treasury note at each month's last close, January 1990 to October 6, 2026, percent; the dashed rule marks the October 6 close.
Against that backdrop, six weeks of the Hang Seng are noisy. The typical six weeks since 1990 moved the index about 5% up or down, and the share of big moves has changed little from one era to the next (Fig. 5). Nearly half of all six-week spans moved the index more than 5.5% either way, close to the 5.4% band that the last 60 sessions’ swings would draw around it today. Any force behind the index has to show itself through that much noise.
In every era, between one six weeks in seven and one in four moved the Hang Seng more than 10% either way.
The median change over 30 sessions, up or down, and the share of 30-session spans that moved the index more than 5.5% and more than 10%, every span from January 1990 to August 2026, by era.
| Years | Typical move, percent | Moved more than 5.5% | Moved more than 10% |
|---|---|---|---|
| 1990 to 2026 | 4.9 | 45% | 20% |
| 1990 to 2007 | 5.5 | 50% | 23% |
| 2008 to 2021 | 4.3 | 39% | 15% |
| 2022 on | 4.9 | 45% | 19% |
Checking your access.
Keep reading for free.
Check your email.
Opening the study.
Leave your email to unlock the full study.
Your membership has ended.
Renew to read all our research in full again.
Renew