Foreign state and institutional holdings of US Treasuries since the escalation of the Iranian conflict

We use exclusively verifiable public data, collected and processed by our Steelldy Risk Engine 3.4 infrastructure and Bloomberg Intelligence modules:

– Treasury International Capital (TIC): Monthly and annual reports, raw and adjusted data for valuation and hedging effects, broken down by country and maturity (source: Treasury.gov).

– Official reserve flows: IMF COFER (Currency Composition of Official Foreign Exchange Reserves), quarterly data.

– Geopolitical risk: Caldara & Iacoviello GPR index (daily update, news-based), aggregated to monthly frequency.

– Treasury microstructure: ATS (Alternative Trading Systems) volumes for blocks > $500M from Bloomberg and Finra TRACE data; dark/ lit volume ratio.

– Composite sanctions risk index: Constructed from 5-year sovereign CDS spreads (China, Saudi Arabia, UAE, etc.), frequency of terms “freeze”, “confiscation”, “sanctions on reserves” in official speeches (NLP extraction via Steelldy Behavioral Matrix 2.1 calibrated on Fed/Treasury/G7 transcripts), and the BIS “weaponization of finance” sub-index.

– Satellite imagery and AIS: Planet Labs data and Steelldy RE 3.8 feeds (Strait of Hormuz traffic, physical gold storage in Gulf and Chinese ports) used as exogenous signals for the dashboard. All series are monthly, from January 2010 to May 2026. Flow variables are expressed in constant billions of USD.

Decoding the Stealth Buyer: A New ATS Signal for Institutional Demand

Foreign demand for US Treasuries is modeled using a revised fundamental equation. The dependent variable is the net foreign holdings of Treasuries, seasonally adjusted and smoothed. Key independent variables include the yield spread between US 10-year Treasuries and a GDP-weighted basket of major reserve currencies, the VIX volatility index, the Geopolitical Risk (GPR) index, an indicator for high ATS block trading volume, net Treasury supply, and a lagged dependent variable for inertia. The model is estimated from January 2010 to May 2026. Results show that a wider yield spread and higher VIX increase foreign demand, while higher GPR reduces it.

The ATS volume indicator is a weak but significant precursor of institutional buying pressure. Net Treasury supply has a negative effect, and the lagged term shows strong persistence. The GPR’s negative impact is concentrated in longer maturities, with a one-standard-deviation increase reducing net purchases by about $20 billion the following month. Structural stability tests (Bai-Perron) find no significant breaks after February 2022, though a TVP-VAR reveals moderate coefficient instability without sign changes.

A complementary panel model for major holders (Japan, China, UK, Saudi Arabia, India) shows significant heterogeneity. Japan and the UK exhibit flight-to-quality behavior, while China and Gulf petro-states are more sensitive to geopolitical risks. These heterogeneous coefficients are incorporated into conditional projections.

Gold and Dollar Dynamics: A Statistical Model of Liquidity, Diversification, and Crisis

A recalibrated 3-regime Hidden Markov Model (HMM) was applied to the residuals of the aggregate demand equation.

Regime 1 (Safe Haven/Liquidity Demand; 36% stationary probability, 6.8 month average duration, low volatility) represents positive net purchases favoring long duration.

Regime 2 (Gradual Diversification/Gold-Currency Rotation; 50%, 9.1 months, moderate volatility) involves moderate sales of long duration.

Regime 3 (Stress/Contagion; 14%, 3.6 months, high volatility) entails forced sales, dislocation, and extreme sanctions. The estimated monthly transition matrix (EM algorithm, 2010-2026) shows probabilities: from Regime 1 (0.85 stay, 0.11 to R2, 0.04 to R3); from Regime 2 (0.06 to R1, 0.88 stay, 0.06 to R3); from Regime 3 (0.03 to R1, 0.12 to R2, 0.85 stay). The probability of transitioning from a normal regime to Stress is 4% (from R1) and 6% (from R2), consistent with historical crises (2008, 2020). Current TIC data shows no activation of Regime 3 since the start of the Iranian conflict. The current signal (June 2026) from the smoothed filter indicates posterior probabilities of 72% for Regime 2, 24% for Regime 1, and 4% for Regime 3. Long duration demand is moderate, offset by strong bill accumulation.

Stress Test Simulates $200 Billion Treasury Sale: 35 bps Yield Jump, Quick Fed Absorption

This paper examines volatility and sanction risk using an augmented GARCH(1,1) model. The model specification is: σ²t = ω + αε²{t-1} + βσ²_{t-1} + γΔGPR_t + δ·SanctionIndex_t. The SanctionIndex is a standardized composite index (3-month moving average of sovereign CDS of major holders, NLP threat score from official speeches, and BIS “weaponization” component). It captures the perceived risk of future freezes or confiscations but remains moderate (current value: 0.28 on a 0-1 scale).

Estimation over 2010-2026 shows: γ = 0.038 (t=2.91), meaning a GPR shock increases conditional volatility by 1.5 annualized percentage points. δ = 0.015 (t=1.96), a marginal effect of sanction risk, not significant at the 5% threshold but retained for stress tests. Compared to the MOVE index, 10-year swaption implied volatility has risen 18% since April 2024, but remains in the high historical range without dislocation. A stress test simulates a partial freeze of 15% of a major holder’s Treasury assets (e.g., China or a GCC country), triggering a forced sale of $200 billion in long-duration holdings over a quarter. The shock is modeled as a point jump in the volatility process (Merton model). Immediate impact on the 10-year yield is +35 basis points, followed by partial normalization (due to Fed and money market fund buying). Conditional VaR for a portfolio with 20% long-duration exposure increases by 2.8%, which is well absorbed by revised allocation (limited tactical duration).