Selected component
U.S. banks
$1.32tn in nominal market value at 2024Q4.
- Data source
- Call Reports
- Frequency
- Quarterly
- Coverage
- 1976Q1–2024Q4
The website harmonizes the source panels to a common quarterly sample from 2011Q4 through 2024Q4.
Data & methods
What the analysis can and cannot say.
Use these definitions and limitations to distinguish observed holdings, estimated responses, and model counterfactuals.
Holdings data
One Treasury market, several source panels.
The holdings dataset combines reporting systems tailored to different investors. The chart shows each sector’s share of the paper’s Treasury universe; select a component to see its source, frequency, and coverage.
Holdings composition · 2024Q4
Treasury holdings by investor sector
Select a component to see the underlying data source.
Source: Dissecting Treasury Market Resilience, Appendix Table A1, “Data sources.” Holdings shares use nominal market values at 2024Q4.
Evidence types
Observed data, estimated responses, and counterfactuals
Observed holdings
Reported sector positions, Treasury supply, yields, and macroeconomic conditions. Holdings accounting describes where debt is held, not why an investor chose it.
Estimated response
Demand functions relating holdings to yields and economic conditions, estimated with instruments and controls under the papers’ assumptions.
Model counterfactual
A re-solved equilibrium after a specified change in supply, investor demand, shock persistence, or policy. It is a conditional model result, not a forecast.
Core data
The data track investors across maturities and over time.
The sector data cover roughly 80 percent of marketable Treasury holdings. The foundational demand estimates use 2011Q4–2022Q4. The resilience analysis updates the sector and country evidence through 2024Q4.
Holdings are organized into short-, intermediate-, and long-term Treasury buckets. The model maps them to representative maturities and uses an affine yield curve for other points.
Country-level foreign holdings use Treasury International Capital reporting. TIC 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. Country labels reflect residence and custody and need not identify the ultimate nationality of the owner.
Interpretation boundaries
Limits of the analysis
What does “resilience” mean here?
Resilience is the inverse of the absolute model-implied yield response to a specified shock. A smaller yield response means greater resilience. The measure does not directly capture bid-ask spreads, trading depth, repo stress, nonlinear market breakdowns, or crisis probability.
Does the debt scenario measure fiscal sustainability?
No. It studies market absorption and the required yield change. It does not calculate the maximum sustainable debt level or default risk.
Is the inflation experiment inflation in isolation?
No. It is an identified cost-push disturbance in which inflation, the policy rate, activity, debt relative to GDP, and relative safe-asset supply move together. The proxy-SVAR provides adequate but not strong joint identification, so the results are conditional model responses rather than precisely identified causal effects.
Are the foreign withdrawals observed sales?
No. For each of the 34 reported locations with at least $10 billion in steady-state holdings, the standardized exercise proportionally scales down the full demand curve so that baseline holdings fall by $10 billion. The rankings use point-estimated location-level demand schedules. Partial pooling attenuates the dispersion but does not reverse the main ranking; formal joint inference on the rankings is not reported.
How should I read the QT experiment?
Weaker expected Fed support, called “rule erosion” in the paper, proportionally scales down the Fed’s estimated yield and macroeconomic loadings. It is a policy counterfactual, not an estimate of investors’ historical beliefs or of separately chosen changes in each loading.
Does the Japan exercise reconstruct the July 31 intervention?
No. The joint U.S.–Japan operation motivates the question, but the table does not estimate its actual financing or market effect. Japan’s separately disclosed April–May intervention amount sets the scale of four counterfactual financing routes. The exercise ranks their direct Treasury-market yield effects, not all economic costs of intervention.