Pressure 03 · Foreign investors

What if foreign investors pull back?

Equal $10 billion contractions produce different 10-year yield responses across reported locations because their maturity holdings and estimated demand schedules differ.

Headline finding

Holdings are global, but equal withdrawals have unequal effects.

Begin with average holdings across all reported locations. The counterfactual tabs then narrow to the 34 reported locations that can support a full $10 billion contraction.

Observed holdings and model counterfactuals

Global Treasury holdings and the locations used in the demand experiment

83 locations shown. This descriptive view includes every separately reported country or jurisdiction in the source panel.

Canada: Average holdings 137.3Kazakhstan: Average holdings 19.4Indonesia: Average holdings 23.3Argentina: Average holdings 4.4Chile: Average holdings 31.2Kenya: Average holdings 0.6Haiti: Average holdings 0.2Dominican Republic: Average holdings 1.6Russia: Average holdings 58.4Bahamas: Average holdings 16.7Norway: Average holdings 88.8South Africa: Average holdings 14.6Mexico: Average holdings 58.3Uruguay: Average holdings 10.8Brazil: Average holdings 254.5Bolivia: Average holdings 0.4Peru: Average holdings 20.7Colombia: Average holdings 33.4Panama: Average holdings 3.3Costa Rica: Average holdings 2.1Honduras: Average holdings 1.5El Salvador: Average holdings 0.6Guatemala: Average holdings 5.2Belize: Average holdings 0.1Venezuela: Average holdings 1.2France: Average holdings 126.8Ecuador: Average holdings 0.5Jamaica: Average holdings 0.5Liberia: Average holdings 0.2Israel: Average holdings 39.4Algeria: Average holdings 1.1United Arab Emirates: Average holdings 55.1Kuwait: Average holdings 37.8Iraq: Average holdings 34.3Oman: Average holdings 13.5Thailand: Average holdings 61.3South Korea: Average holdings 94.0India: Average holdings 148.0Pakistan: Average holdings 1.8Sweden: Average holdings 40.9Ukraine: Average holdings 4.9Poland: Average holdings 29.8Austria: Average holdings 4.6Hungary: Average holdings 1.4Romania: Average holdings 5.3Germany: Average holdings 78.1Greece: Average holdings 3.2Turkey: Average holdings 32.0Switzerland: Average holdings 237.1Luxembourg: Average holdings 120.5Belgium: Average holdings 111.6Netherlands: Average holdings 52.5Portugal: Average holdings 4.3Spain: Average holdings 36.2Ireland: Average holdings 253.1New Zealand: Average holdings 5.7Australia: Average holdings 41.6China: Average holdings 1094.6Taiwan: Average holdings 203.5Italy: Average holdings 38.3Denmark: Average holdings 16.1United Kingdom: Average holdings 288.1Philippines: Average holdings 42.1Malaysia: Average holdings 12.7Finland: Average holdings 6.4Czech Republic: Average holdings 5.9Japan: Average holdings 1147.3Paraguay: Average holdings 1.6Saudi Arabia: Average holdings 120.8Cyprus: Average holdings 0.2Morocco: Average holdings 2.6Egypt: Average holdings 4.0Trinidad and Tobago: Average holdings 3.6Bermuda: Average holdings 73.1Cayman Islands: Average holdings 233.2Hong Kong: Average holdings 195.3Singapore: Average holdings 139.2British Virgin Islands: Average holdings 39.8Barbados: Average holdings 2.9Netherlands Antilles: Average holdings 1.8Bahrain: Average holdings 1.1Monaco: Average holdings 0.2Aruba: Average holdings 0.2
Colored areas and markers show all 83 separately reported countries and jurisdictions. Holdings are quarterly averages over available observations in 2011Q4–2024Q4.
Model counterfactual

Interpretation: The holdings tab is descriptive and covers 83 separately reported countries and jurisdictions. The counterfactual tabs isolate differences in maturity holdings and estimated demand schedules by holding the withdrawal fixed.

Source: Dissecting Treasury Market Resilience, Figure 3 and Appendix Tables A5–A6. Counterfactuals are point estimates for 34 sufficiently large locations; TIC labels reflect reporting residence and custody rather than final beneficial ownership.

Economic interpretation

Size alone does not predict market impact.

The model proportionally scales down each reported location’s full demand curve so that baseline holdings fall by $10 billion. Canada and India provide a sharp comparison: their yield effects differ because their estimated demand schedules differ, even when the dollar amount is held fixed.

Treasury International Capital reports bills and longer-term Treasuries separately. The analysis scales reported bill holdings to represent the under-one-year bucket, then divides longer-term holdings between the intermediate and long buckets using aggregate TIC maturity shares.

Model counterfactual

How to interpret the experiment

The experiment is a permanent proportional scaling of a reported location’s full demand curve. It does not identify the nationality of the final beneficial owner.

Live policy case

Why did the U.S. Treasury join Japan’s yen intervention?

On July 31, 2026, Japan’s Ministry of Finance and the U.S. Treasury carried out their first coordinated yen-buying intervention since 1998. The yen had weakened beyond ¥163 per dollar, near a four-decade low. Japanese households and businesses faced higher import costs, while officials in both countries were concerned that the speed and disorderliness of the decline could spread instability beyond the currency market.

Actual eventThe U.S. Treasury bought yen

The New York Fed, acting for the Treasury, reportedly sold euros and bought yen. The U.S. leg therefore supported the yen without selling U.S. Treasuries.

Policy optionJapan needs foreign currency to buy yen

Japan can sell reserve assets, including Treasuries, to raise dollars. It also plans to use the Fed’s FIMA Repo Facility, which would let it borrow dollars against Treasuries without selling them.

Research questionWhat must the private market absorb?

If Treasuries are sold, investors must take them onto their balance sheets. The yield effect depends on how much duration reaches the market and whether the sale is expected to reverse.

Observed scale

$73.6 billion of yen support

Before the joint operation, Japan disclosed ¥11.7349 trillion, about $73.6 billion, of yen intervention between April 28 and May 27, 2026. We use that amount to compare four possible financing routes. The comparison does not reconstruct the July 31 trade. The release does not report that Japan sold $73.6 billion of Treasuries.

Model counterfactual
Model-implied 10-year Treasury yield effects at the scale of Japan’s disclosed intervention
Financing routeWhat reaches the Treasury market10-year effect
FIMA repo financingNo Treasuries reach the market0.00 bp
U.S. bill issuanceShort-duration supply reaches the market+0.49 bp
Temporary Japanese saleLonger duration, expected to reverse+2.62 bp
Permanent retrenchmentLonger duration, demand does not return+4.23 bp

Interpretation: The July 31 joint intervention and Japan’s earlier $73.6 billion intervention amount are observed. The financing routes in the table are counterfactuals. Repo’s zero is a direct Treasury-market effect by construction, not a statement about the total economic costs of intervention.

Source: Ministry of Finance, Japan, August 3, 2026; Associated Press, August 3, 2026; Axios, August 3, 2026; Foreign Exchange Intervention Operations, May 29, 2026; Dissecting Treasury Market Resilience, Table 2 and Section 3.6.