Public-Data Research Monitor
Monitoring how the Supplementary Leverage Ratio shapes large-bank Treasury holdings and balance-sheet behavior—using free public regulatory data from FFIEC Call Reports, FR Y-9C, FR Y-15, NY Fed dealer statistics, and FINRA TRACE.
Core Results
The flagship design compares balance-sheet-constrained banks to less-constrained peers around the April 2020 temporary SLR exclusion. Constrained banks—especially covered-bank subsidiaries—appear to have increased Treasury holdings more than comparison banks.
The low_headroom_treated and
covered_bank_treated coefficients are positive and
economically meaningful across all three sample variants. In the
broad sample, both are significant at the 1% level.
The high_ust_share_treated coefficient is near zero
and insignificant everywhere. The flagship-clustered results lose
conventional significance, so the headline should be framed as
suggestive rather than definitive.
Research Design
A panel event study around the April 2020 temporary SLR exclusion, using difference-in-differences with entity and time fixed effects.
The Supplementary Leverage Ratio requires large banks to hold Tier 1 capital against all on-balance-sheet assets and certain off-balance-sheet exposures, without risk-weighting. Unlike risk-based capital ratios, SLR treats Treasuries the same as corporate loans in the denominator. For banks near the binding constraint, every dollar of Treasuries consumes scarce leverage capacity—creating a direct link between capital regulation and Treasury-market intermediation.
GSIB parent holding companies face a 5% enhanced SLR (eSLR) minimum, while
their covered insured-bank subsidiaries must maintain 6% to be classified as
well-capitalized. The 2025 final rule recalibrates these buffers to
3% + 0.5 × method 1 surcharge for parents and
3% + min(1pp, 0.5 × parent surcharge) for subsidiaries,
effective April 2026.
On April 1, 2020, the Federal Reserve temporarily excluded U.S. Treasury securities and deposits at Federal Reserve Banks from the SLR denominator for holding companies. The exclusion expired on March 31, 2021. This created a clean before/during/after window for estimating how leverage-ratio relief affects bank Treasury demand.
The event window runs from 2019–Q1 through 2021–Q4, with the pre-period ending at 2020–Q1 and the post-treatment period beginning at 2020–Q2. This gives four pre-treatment quarters and seven post-treatment quarters.
The universe covers U.S. GSIB parents, major insured-bank subsidiaries, large non-GSIB controls, and U.S. IHCs of foreign banking organizations. The broad sample includes 21 entities (232 bank-quarter observations). The flagship per-parent sample restricts to the single largest insured subsidiary per parent family—15 entities, 15 parent clusters, 180 observations—to avoid double-counting families with multiple large subsidiaries.
Three treatment arms, each splitting the sample on a pre-shock characteristic measured as of 2019–Q4:
The baseline is a two-way fixed-effects difference-in-differences model:
Yit = αi + γt + β · (Treatedi × Postt) + εit
Entity fixed effects (αi) absorb time-invariant bank characteristics; quarter fixed effects (γt) absorb common macro shocks including aggregate market conditions. The coefficient of interest β measures the differential change in outcomes for treated banks after the exclusion. Standard errors use HC1 heteroskedasticity-robust inference in the baseline, with parent-family clustering as the strictest check.
All outcomes are drawn from FFIEC Call Report schedules and scaled by total assets to ensure comparability across bank sizes:
Quarter-level market conditions—NY Fed primary-dealer positions, UST repo volumes, and TRACE aggregate par value—are absorbed by the quarter fixed effects in the baseline specification. The repo additionally reports an interaction sensitivity (Treatedi × Postt × Markett) and a weaker auxiliary specification without time fixed effects that lets raw market levels enter directly. Neither overturns the baseline read.
Mechanism Evidence
Five current extension reports build a mechanism story: reallocation away from other balance-sheet uses, a shift in safe-asset composition toward Treasuries, consistency with family-level transmission, and a possible reduction in trading-balance-sheet use.
The Treasury response came alongside weaker expansion elsewhere on the balance sheet. Constrained banks appear to reallocate scarce capacity rather than simply growing everything.
| Treasury Inventory | Fed Balances | Deposit Growth | Loan Growth | |
|---|---|---|---|---|
| Low Headroom | +2.47pp | +0.26pp | −3.52pp | −1.65pp |
| Covered Bank | +1.97pp | −0.45pp | −3.98pp | −2.32pp |
The main difference is composition within safe assets, not just the level of Fed balances. Constrained banks shifted the safe-asset mix toward Treasuries by roughly 9–10 percentage points more than controls.
Constrained banks reduced trading-balance-sheet use while absorbing more Treasuries. The strongest market linkages are between treated-minus-control trading-assets gaps and NY Fed dealer-position context (correlation up to 0.82).
The insured-bank Treasury result sits inside broader family-level balance-sheet behavior. Linked-family analysis covers 100 bank-parent quarter observations across 15 parent families.
High-surcharge families show lower average bank headroom (0.87pp vs. 1.90pp), consistent with tighter SLR constraints driving more pronounced balance-sheet responses.
Broader View
A longer quarterly panel places the 2020 event in context. The temporary exclusion window stands out most clearly in reserve accumulation, while Treasury share rose modestly. Over the subsequent QT era, parent trading-assets share declined as dealer positions and TRACE par volume grew.
Extension
The 2020 event study isolates the SLR channel. But leverage headroom is not the only pressure that shapes bank Treasury capacity. This module decomposes balance-sheet constraints into leverage, duration-loss, and funding pressure across insured-bank and parent/IHC panels through 2025Q4.
In the 2022–2023 duration-loss window, duration pressure becomes the dominant constraint for both insured banks (65.9%) and parents/IHCs (63.0%). By late QT normalization, insured banks still lean duration loss at 42.1%, while parents shift back toward leverage at 35.4%.
Linked families match on the dominant constraint in 64.8% of duration-loss-window family-quarters, and both sides are duration-loss dominated in 48.6% of those observations. The 2022–2023 regime is the first period where joint duration-loss dominance becomes common.
On the parent panel, higher duration pressure is associated with higher Treasury share in the 2022–2023 window (0.027, p = 0.001) and lower Treasury share in late QT (−0.034, p = 0.026). Bank-side results are weaker. This is supporting evidence, not a separate causal claim.
Public Data
All data comes from free public regulatory filings and market-data APIs. No proprietary feeds, no paywalled datasets.
The main empirical panel. Bulk schedule files provide bank-level total leverage exposure, Tier 1 capital, Treasury holdings, Fed balances, repos, trading assets, loans, and deposits.
Parent-level Tier 1 capital, total leverage exposure, Treasury holdings, and trading assets. Historical files from the Chicago Fed archive; current data via NIC automation.
Method 1 surcharge context and GSIB systemic-intensity overlays. Snapshot files parsed via Playwright; surcharge scores from the OFR Basel workbook.
Weekly aggregate data on dealer Treasury positions, transactions, repo, reverse repo, and securities lending. Collapsed to quarterly market-context overlay via the official Markets Data API.
Free public aggregate data on Treasury par volume and trade counts. Weekly archive files for 2019–2022; monthly workbooks for 2023+. Treated as market-level context only.
Bank Systemic Risk Monitor scores used to build the annual Basel-method surcharge overlay. Provides ranking context and systemic-intensity data for GSIB identification.
Automated downloaders fetch Call Report bulk ZIPs, FR Y-9C files, FR Y-15 snapshots, NY Fed API data, and TRACE aggregates into standardized raw storage.
Raw files are extracted, merged by schedule, and normalized into Parquet with variable lineage and quality checks.
A curated crosswalk links entities across sources. Insured-bank and parent panels are assembled with rule-aware SLR headroom, Treasury-capacity metrics, and market overlays.
The event-study workflow runs DiD estimates, event-time coefficients, and extension reports. All outputs are written to reproducible report directories.
Output
The repo generates structured report outputs covering the core event study, five current extension analyses, market context, and a live constraint-decomposition module.
Integrated summary of all results, caveats, and recommended framing for the v1 release.
Core DiD estimates and event-time coefficients for the temporary exclusion window. Includes broad, flagship, and flagship-clustered variants.
Treated-versus-control change decomposition across Treasury, Fed balances, deposits, and loans.
Composition analysis separating Treasuries from Fed reserve balances within the safe-asset portfolio.
Linked-family analysis comparing bank-subsidiary and parent-company balance-sheet changes within the same corporate family.
Sensitivity analysis linking trading-assets and Treasury-inventory responses to dealer-position and repo market context.
Broader quarterly panel with regime averages across pre-exclusion, temporary exclusion, post-exclusion normalization, and QT era.
NY Fed dealer statistics and TRACE aggregate data providing Treasury-market backdrop for the bank-level analysis.
Cross-regime comparison of leverage headroom, duration-loss pressure, and funding stress for insured banks and parents through 2025Q4, with family-alignment summaries and a supporting interaction-regression layer.