What if Asia simply bought less?

The decision could fit on a single line in an investment budget: keep the bonds already owned, reinvest maturing principal and put additional money somewhere else. The portfolio remains in place, but the market expecting those new purchases must find another holder. The final chapter of The Asian Dollar Loop follows this quieter choice: the next purchase that never happens.

The previous chapter’s stress test traced the interaction between valuation losses, exchange rates and cash needs. This ninth chapter follows a different path. An insurer or pension fund can change its allocation before it faces pressure to sell. It may judge that the yield left after currency hedging no longer adequately compensates for the risks. Or a domestic bond may have become a better match for its liabilities.

For the US borrower, that raises a practical question. Who will hold the next securities, at the required maturities, and on what terms?

Research for this article closed on 5 September 2026. The evidence does not establish a collective Asian buyers’ strike. It shows demand diverging across asset classes and statistics in which bond redemptions can look like sales. The decisions illustrated below are explicit scenarios. They explain a mechanism without assigning a common intention to an entire region.

Slower demand becomes visible when assets are separated

The Treasury’s TIC release of 17 August 2026 reports $329.3 billion in foreign private net purchases of long-term Treasuries over the twelve months ending in June 2026, down from $561.1 billion over the preceding twelve months. That is a calculated decline of about 41.3%. Demand remains positive, but its scale has changed.

June illustrates the divergence: foreign private investors bought a net $144.7 billion of equities and related instruments, versus $16.6 billion of Treasury notes and bonds. They were net sellers of agency bonds. These are worldwide figures, without a separate Asian insurer category. They nevertheless show why aggregate capital inflows cannot answer a question about demand for government debt.

It would be equally misleading to treat this comparison as evidence of a general rejection of US assets. Buying a US equity and buying a Treasury involve different risks, mandates and holding periods. Money available to one segment can diminish while another attracts more.

This chapter is about that selectivity. An institution can retain substantial US exposure while reducing the share of its next investments allocated to long-dated dollar bonds. It can shorten duration, demand more compensation for credit risk or hold more cash. A ranking of the largest foreign holders captures these decisions poorly.

Japan’s negative net flow still contains substantial buying

Japan’s investor-type statistics narrow the investigation to life insurers. From January through July 2026, reporting life insurance companies acquired ¥5,719.7 billion of foreign long-term debt securities. The equivalent figure for the same months of 2025 was ¥5,824.2 billion. Gross purchases fell relatively little.

Dispositions rose from ¥6,252.5 billion to ¥7,050.9 billion. The sum of published monthly net transactions consequently reached minus ¥1,331.1 billion in 2026, compared with minus ¥428.4 billion a year earlier. Small differences from subtracting the rounded gross totals reflect rounding in the published series.

The essential word is “dispositions”. Japan’s Ministry of Finance explains that this category includes bonds redeemed at maturity. An insurer can therefore record a net disposition without selling a single security before it matures. The series also covers foreign long-term bonds across destinations and currencies. It is not a Treasury-only measure.

Those definitions change the conclusion. A negative net flow is established. A discretionary liquidation of US assets motivated by distrust is not. And “they have stopped buying” describes poorly a sector that still made trillions of yen in acquisitions over seven months.

An individual institution’s accounts offer another useful check. In Nippon Life’s general account, reported foreign bonds increased from ¥12,252.1 billion on 31 March 2026 to ¥12,526.1 billion on 30 June. Their reported weight remained 14.5%. The stock rose. That does not establish an increase in net purchases: prices, currencies, redemptions and transactions still need to be separated. It does rule out presenting the sector’s negative net flow as a uniform decline across every institution’s reported holdings.

Stocks and flows answer different questions. One records the securities present on a date; the other records transactions during a period. Moving between them requires a decomposition, as the Federal Reserve’s work on the CSLT dataset explains. A growing balance sheet can contain little fresh money. A shrinking one can still contain substantial purchases.

Three decisions hidden inside “stop buying”

Consider a hypothetical portfolio worth 400 billion dollars, with 40 billion dollars of principal maturing during the year. Its owner had planned to invest a further 20 billion dollars of new money. These figures represent no particular country or insurer.

The first policy is to continue the programme. Reinvest all 40 of maturities and add the planned 20. Gross purchases are 60; the closing stock is 420, with prices and exchange rates held constant.

The second policy is to freeze new money. Reinvest all maturing principal, but direct the additional 20 elsewhere. Gross purchases are still 40. The closing portfolio remains 400. Holdings are unchanged, yet the market receives 20 less than under the original plan.

The third policy is to reinvest only half the maturing principal and cancel the additional allocation. Gross purchases fall to 20 and the closing stock to 380. The portfolio shrinks through scheduled redemptions. No sale before maturity is required.

Stopping every purchase is a fourth, stronger policy: no reinvestment and no new money. The stock in this example would fall to 360 after one year. That is materially different from simply freezing additional allocations. Conflating the two makes the scenario look more severe than it is.

The distinction also matters when describing flows. Reducing planned portfolio growth from 20 to zero removes 20 of demand relative to the original programme. It does not mean selling 20 on the market. The relevant amount depends on the counterfactual: the purchases that would have occurred if the institution had maintained its policy.

Change new money and maturity reinvestment

The tool below extends this principal-flow accounting over one to five years. It holds the redemption schedule constant across policies to make the comparison transparent. Newly purchased bonds mature beyond the selected horizon. Prices, coupons, exchange rates and defaults are excluded.

LAB / THE NEXT PURCHASE

Keep the portfolio. Cut the purchases.

Change new allocations and maturity reinvestment. All amounts are hypothetical, in USD billions.

Keep addingFreeze new moneyReinvest one halfStop all purchases

Scenario assumptions

Opening portfolio

400

USD bnPlanned annual new money

20

USD bnShare of new money still invested

0

%Share of maturities reinvested

100

%

Maturities and horizon

Opening principal repaid each year

10

%Horizon

3

years

Reset

Gross purchases per year

40USD bn

Net flow per year

0USD bn

Closing stock

400USD bn

Cumulative purchases forgone

60USD bn

Relative to continuing the original programme

Annual flows

Redemptions

40

Of which reinvested

40

Allocated new money

0

Stocks and purchases forgone, in USD billions

Year	Keep adding	Scenario	Purchases forgone
Start	400	400	0
1	420	400	20
2	440	400	40
3	460	400	60

Purchases = reinvested maturities + allocated new money. Net flow = purchases − redemptions.

Purchases forgone measure the reduction in this demand against the original programme. They are neither a valuation loss nor a forecast of higher yields.

No sales before maturity. Annual redemptions come from the opening portfolio; newly purchased bonds mature beyond the chosen horizon. Prices, FX, coupons and defaults are excluded.

Under the initial settings, the stock remains 400 for three years. Continuing the original programme would have taken it to 460. The gap of 60 billion dollars is cumulative purchases forgone, rather than a loss on the securities retained. Reinvesting only half of maturities leaves a closing stock of 340 and a 120 gap relative to the original plan.

The tool does not identify where the money goes next. It might fund a Japanese bond, a European company, another US asset or a payment to a beneficiary. Nor does it convert each dollar of purchases forgone into a basis point increase in yields. That would require an estimate of how other buyers respond.

The next allocation can change before the portfolio does

An insurer’s portfolio carries the history of its policies, past yields and accounting decisions. Its next investment is made at the terms available now. That difference allows a gradual transition: retain the legacy assets while redirecting incremental flows.

The Bank of Japan’s April 2026 Financial System Report describes caution about accumulating unhedged foreign bonds. It also notes insurers’ caution toward super-long Japanese government bonds amid heightened volatility. A domestic alternative exists, but a higher yield does not make every maturity attractive.

Returns need to be compared in the currency of the liabilities and against comparable risks. In a purely illustrative example, a foreign bond yielding 5%, with an annualised currency hedge cost of 2%, leaves roughly 3% of carry before other costs and risks. A domestic bond yielding 3.2% could then compete. These are teaching assumptions rather than current market quotes.

The subtraction is only a first screen. A short-dated hedge rolled repeatedly does not lock in the same cost over a long bond’s lifetime. Forward pricing, interest differentials, intermediary balance sheets and collateral requirements all matter. BIS research on portfolio investments and FX derivatives explains why this financial layer is part of the investment decision itself.

Taiwan faces a different constraint. In a December 2025 explanation, the Financial Supervisory Commission linked insurers’ foreign bond investments to the domestic market’s insufficient capacity and yields relative to their long-term obligations. Bringing money home requires assets capable of absorbing it. A desire to reduce currency risk does not create those assets.

Pension funds operate under yet another framework. The GPIF policy portfolio effective from April 2025 retains four 25% targets: domestic bonds, foreign bonds, domestic equities and foreign equities. This structure governs diversification and rebalancing. It neither describes an entirely US portfolio nor translates into a daily decision to buy Treasuries.

There is no single investment instruction for “Asia”. A Taiwanese insurer, GPIF and a Chinese reserve manager have different liabilities and mandates. The chapter on China traced that separation of balance sheets. Transactions that look alike in an aggregate dataset can arise from different decisions.

Treasury still has to finance the next flow

On 3 August 2026, the US Treasury projected $739 billion of privately held net marketable borrowing for July–September, assuming a $950 billion end-September cash balance. This is a forecast. It is neither the quarterly fiscal deficit nor gross auction issuance, which would include refinancing.

The figure provides a sense of scale, without supplying a ratio for this investigation. Seven months of Japanese transactions in yen, across foreign issuers, cannot be divided by one quarter of US borrowing in dollars to measure a “withdrawn financing share”. The currencies, periods and instruments do not match.

The economic relationship is straightforward. If some investors reduce their net acquisitions while the amount of debt to be held keeps increasing, other investors must absorb more. That can happen through auctions, secondary-market transactions or intermediary inventories. Buying an outstanding bond can free the seller’s capacity to subscribe to a new issue.

A marginal withdrawal therefore does not require an empty auction. The market can keep functioning with different buyers and less favourable terms for the borrower. A possible symptom is a yield higher than it would have been if the earlier demand had persisted. Establishing that effect requires separating it from other drivers of interest rates.

Maturity matters too. Money available for a three-month bill does not automatically replace an investor willing to hold twenty-year debt. Issuing shorter securities can attract a different clientele, but brings the next refinancing closer. The amount of financing absorbed and the duration of the risk absorbed are separate dimensions.

The replacement buyer may have a different business model

Treasury’s annual survey offers a useful precedent. Between June 2024 and June 2025, foreign private investors made about $570 billion of net purchases of long-term Treasuries, while foreign official investors recorded approximately $100 billion of net sales. Table 1 of the survey shows investor categories moving in opposite directions. It does not guarantee that the same substitution will recur at unchanged prices.

A bond manager may increase exposure once yields become attractive. A bank may want liquid securities, subject to capital and interest-rate constraints. A money market fund primarily serves the short end. A dealer may warehouse securities temporarily. An arbitrage fund may hold cash bonds against futures positions, financed through repo. These investors do not offer interchangeable capacity to hold risk over time.

The OECD’s Global Debt Report 2026 highlights the shift toward more price-sensitive investors and, in some segments, greater reliance on leverage. Such investors help absorb changes in supply and demand. Their presence can also make markets sensitive to the financing conditions supporting their positions.

Geography creates a further problem. In an October 2025 research note, Federal Reserve researchers show how Treasury positions held by Cayman-domiciled funds can be missed in TIC statistics when securities circulate as repo collateral. Treating each custody or domicile location as a homogeneous saving country would be unsafe. So would adding historical research adjustments to the latest data without checking subsequent revisions.

The relevant question is whether the replacement investor can retain the risk when prices or funding conditions become less favourable. Finding a buyer today and finding the same buyer tomorrow are separate requirements.

Neither yields nor the dollar deliver an automatic verdict

Less demand for long bonds can put downward pressure on prices and upward pressure on yields, other things equal. But other things change. Expected short rates can fall, growth can weaken, inflation can surprise and new buyers can return at the available yields. The observed interest rate reflects these forces together.

The New York Fed’s framework separates expected short-term interest rates from an estimated term premium. An increase in that premium does not identify who caused it. A negative Asian flow number released around the same time cannot establish attribution.

Economic research underscores the importance of responses. A Federal Reserve study published in 2012 estimates different effects from a shock to foreign official purchases depending on whether private investors are allowed to respond to the change in yields. Its historical sample and official-sector buyers do not calibrate insurers in 2026. The transferable lesson is the substitution mechanism, rather than a coefficient to paste into a calculator.

Currency exposure requires another distinction. Investors can keep their bonds while increasing hedges, selling more dollars forward. BIS Bulletin 105 examines that mechanism during April and May 2025. Conversely, allowing a Treasury to mature and moving the proceeds into another dollar asset need not involve an immediate conversion into local currency. Asset allocation and currency allocation have to be tracked separately.

Nor does the macroeconomic loop disappear because a fund changes its mandate. The balance of payments links current transactions and financial transactions. At the aggregate level, an external surplus corresponds to net acquisition of assets or reduction of liabilities vis-à-vis the rest of the world, subject to other accounts and adjustments. That identity does not require any particular insurer to buy a Treasury. Flows can pass through different institutions, instruments and countries; exchange rates, yields and spending can also adjust.

How to identify a gradual withdrawal

A convincing case would require several observations to line up over comparable periods. A monthly holder ranking or one weekly net-flow figure cannot do that work alone.

• Gross purchases by investor type. Are insurers’ acquisitions declining persistently? The Japanese series supports that investigation within its designated-reporter and foreign-bond coverage.
• Maturities and reinvestment. Does a shrinking stock reflect bonds running off, discretionary sales or prices? Institutional maturity schedules and reports must supplement the aggregates.
• Returns available after hedging. Compare similar instruments, maturities and risks, including the cost of rolling protection. A spread between two government bond yields is insufficient.
• The destination and duration of new allocations. Moving into short-term US paper differs from returning to domestic bonds. Mandates and allocation reports need to identify the change.
• Other buyers’ capacity. Holdings composition, inventories and repo conditions help identify who absorbs the risk. Series must be reconciled without counting the same securities twice.
• Prices and competing explanations. Persistent pressure on particular segments would be consistent with weaker demand. Attribution requires controlling for supply, expected monetary policy and other market news.

The evidence collected here establishes divergent demand and the need to distinguish its mechanisms. It does not reconstruct a comparable, institution-by-institution Asian reinvestment rate. The tool remains an accounting exercise, without an implicit estimate of that rate hidden inside it.

This is where the series ends. The risk can begin before the forced sale explored in the stress test: in the terms an insurer requires to commit the next yen, euro or dollar of its policyholders’ money. It can also recede if available yields, hedging costs or diversification needs bring buyers back.

Asia can retain much of its portfolio while contributing less to future purchases. For the United States, the possible change lies in the price demanded by the next holder and that holder’s ability to stay. Existing portfolios are visible. The marginal decision requires opening the flows.