Skip to content
FinObservatory

Short-term funding / Methodology

US Short-Term Funding Volumes: methodology

This page documents /funding: where every series comes from, what each exhibit plots, the exact API endpoints and mnemonics, the vintage and gap conventions, the build-time verification, and the license under which the data is republished.

What this layer is, and is not

/funding is the volume side of US money markets: how many dollars move through the repo market daily and where money-market funds place their cash monthly. It deliberately carries no rates. The money-market rates (SOFR, EFFR, the FOMC target range, the SOFR averages) live on /conditions, built from the NY Fed Markets Data API. The two pages are complements: /conditions prices the plumbing, /funding measures its flow. Every series here is a market-level aggregate statistic exactly as published by the OFR; the API serves no counterparty-level or fund-level microdata and none is carried.

Source

Office of Financial Research (OFR), U.S. Department of the Treasury, Short-Term Funding Monitor (https://www.financialresearch.gov/short-term-funding-monitor/), via the OFR's free, no-authentication REST API:

  • https://data.financialresearch.gov/v1/metadata/mnemonics: the full series universe (442 mnemonics across 5 datasets at the 2026-07-17 pull).
  • https://data.financialresearch.gov/v1/metadata/query?mnemonic=<m>: name, description, unit, frequency, start date, release, and series structure.
  • https://data.financialresearch.gov/v1/series/full?mnemonic=<m>: the full observation history. This is the primary data endpoint.
  • https://data.financialresearch.gov/v1/series/timeseries?mnemonic=<m>: an independent endpoint over the same series, fetched at build time as a cross-check (must match the primary endpoint exactly on every non-null observation, for three series).

The two underlying data releases, both served through the monitor:

The OFR Financial Stress Index is not on this API; FinObservatory carries it separately (on /conditions) from the OFR's static FSI files. That route-of-record finding is documented in scripts/fetch_ofr.py.

The curated series

Twenty series, fetched in full and rebuilt idempotently by scripts/build_ofr_stfm.py into the ofr_stfm parquet (long format: mnemonic, date, value in whole USD) plus the embedded catalog ofr_stfm_meta (label, OFR name, unit, frequency, start date, release). The series table rendered at the bottom of /funding is queried from that catalog joined to the data itself, so mnemonics, start dates, and latest observations on the page always reflect what the API actually served at the last refresh.

Exhibit-to-mnemonic map:

ExhibitOFR mnemonics
Latest-readings stripREPO-TRI_TV_TOT-P, REPO-DVP_TV_TOT-P, REPO-GCF_TV_TOT-P, REPO-DVP_OV_TOT-P, MMF-MMF_TOT-M, MMF-MMF_RP_TOT-M
Repo transaction volume by segmentREPO-TRI_TV_TOT-P, REPO-DVP_TV_TOT-P, REPO-GCF_TV_TOT-P
Tri-party volume by collateral classREPO-TRI_TV_T-P, REPO-TRI_TV_AG-P, REPO-TRI_TV_CORD-P, REPO-TRI_TV_O-P
MMF investments by asset classMMF-MMF_TOT-M, MMF-MMF_T_TOT-M, MMF-MMF_AG_TOT-M, MMF-MMF_BRA_TOT-M, MMF-MMF_OA_TOT-M, MMF-MMF_RP_TOT-M
MMF repo lending by counterpartyMMF-MMF_RP_wFR-M, MMF-MMF_RP_wFICC-M, MMF-MMF_RP_wDFI-M, MMF-MMF_RP_wFFI-M, MMF-MMF_RP_wOCP-M

Also acquired and shown in the strip and catalog: REPO-GCF_OV_TOT-P (GCF outstanding volume; catalog only for the chart set, strip carries DVP outstanding).

Definitions and conventions

  • Transaction volume (TV) is the volume of repurchase agreements starting on a given day in that service: a daily flow of new trades. Outstanding volume (OV) is the stock of repo open in the service on the day. The two are never mixed on one chart.
  • Tri-party covers repo settled on the tri-party platform (the OFR's description: "transaction volume of all repurchase agreements starting on a given day that were settled in tri-party repo"). DVP and GCF are the Fixed Income Clearing Corporation's delivery-versus-payment and general-collateral-finance services. The API also serves a tri-party companion collection whose metadata name reads "excluding Federal Reserve transactions" (TRIV1 mnemonics); this layer carries the headline TRI collection.
  • Vintage: repo series are the Preliminary (-P) vintage, the one the OFR's own monitor charts. The Final (-F) vintage lags: checked 2026-07-17, the Final tri-party total was last updated 2026-03-31 while the Preliminary was updated same-day. A revision policy caveat follows from this: recent Preliminary observations can revise.
  • Units: the API publishes these series in whole USD (unit metadata USD, magnitude 0, asserted at build). Charts and stat tiles divide by 1e12 and label USD trillions.
  • Gaps are real. The OFR's own series notes state: "Missing values in this series represent observation periods in which either no trading took place or in which disclosure edits were applied to protect business-confidential information." Published nulls are dropped at build (never imputed, never zero-filled), and chart lines break across gaps rather than bridging them.
  • Coverage starts differ by design. Tri-party daily data begin 2014-08-22, FICC DVP/GCF 2018-05-07, the MMF collection 2010-11-30, with some counterparty splits starting later. Lines simply start when their series does.

Build-time verification (all in scripts/build_ofr_stfm.py, all fail loudly)

  • Universe check: every curated mnemonic still served by the API.
  • Metadata checks: unit USD magnitude 0, expected frequency, and the metadata start date equal to the first observation actually served, per series.
  • Cross-endpoint check: /series/timeseries equals /series/full on every non-null observation for three series.
  • Additivity: tri-party total equals the sum of its four collateral classes on every published date; MMF total equals the sum of its five asset classes every month; tolerance $1 on values in the trillions.
  • Anchors from the verified 2026-07-17 pull: tri-party volume on 2019-09-17 (the repo-spike day) of $1,441,206,792,080.00; the first MMF month (2010-11-30) at $3,076,353,570,850.39; and the peak month of MMF repo with the Federal Reserve landing in 2022-12. An anchor drift means the OFR revised history: the build stops so the revision is reviewed, never silently absorbed.
  • Freshness: every daily series within 10 days of the pull, every monthly series within 45 days.
  • Zero duplicate (mnemonic, date) keys, zero nulls, all values positive.

Every number displayed on /funding is computed at build time from the parquet through src/lib/stfm.ts (DuckDB over ofr_stfm / ofr_stfm_meta); nothing is typed into the page. Daily charts are decimated with the site-wide min/max downsampler (src/lib/downsample.ts), which only ever drops rows, so every rendered point is a real observation and daily spikes survive exactly.

Refresh

The layer refreshes on the daily tier of scripts/refresh.sh (see docs/REFRESH_RUNBOOK.md): a full re-fetch of all 20 series, idempotent, with the assertions above re-run on every refresh.

License and citation

This page republishes OFR-served aggregate series. The data is a US federal government work. OFR's legal notice (https://www.financialresearch.gov/legal-notices/) states: "No copyright may be claimed for any work on this website that was created by a federal employee in the course of his or her duties. However, credit is requested if you reproduce or copy any such work." Credit is rendered on /funding and here.

Cite the source as: Office of Financial Research (OFR), U.S. Department of the Treasury. Short-Term Funding Monitor. https://www.financialresearch.gov/short-term-funding-monitor/ (U.S. Repo Markets Data Release; U.S. Money Market Fund Data Release), retrieved via data.financialresearch.gov/v1.