Asset-Backed Securities Distributions (Form 10-D)

Form 10-D distribution reports for asset-backed securities: which trust paid out, for which distribution period, its depositor and sponsor, the Regulation AB asset class, and whether asset-level data rode along with the filing. A monthly census of every publicly registered US securitisation trust and the deal parties behind it.

edgar_abs_distributions - the dataset name to pass to the Obscura API.

What one row means

One **Form 10-D asset-backed issuer distribution report** - the mandated cover-page identity of one distribution period for one securitisation trust. 105,652 filings since 2015: a complete monthly census of every publicly-registered US securitisation, and the only public place where a trust, its depositor, its sponsor and its Reg AB asset class are stated together, per period, as identified CIKs. # What a row IS, and what it deliberately is not A row is the SUBMISSION'S IDENTITY: which trust, which period, which deal parties, which exhibits rode along. It is **not** the pool's numbers. Those live in `EX-99.1`, which Reg AB Item 1121 prescribes the *content* of and not the *shape* of: three opened servicer reports share no row labels, no column headers and not even a table count (10 vs 33 vs 33). A single field parser across 105,652 of those does not exist, so this table does not pretend to one. See `docs/datasets/edgar_abs_distributions.md`. # Every column here is a STRUCTURED SEC header field - none is parsed prose The design proposed reading the depositor, the sponsor and the trust off the cover page's English. It does not have to: SEC's own submission header states `<DEPOSITOR-CIK>`, `<SPONSOR-CIK>` and `ABS ASSET CLASS` outright, measured present on **156 of 156** filings sampled across 2015 and 2026. So the whole row is evidence class B, and the cover-page prose - the distribution date, the period start, the note-class table - is left unread rather than half-parsed. # The row costs 3.5 KB of bronze, not 52 MB The submission this describes is landed by [`edgar_abs_assets`](super::edgar_abs_asset) territory and can be enormous: one dissemination day of 10-D (2026-06-15, 106 filings) is **5.39 GB** of full submissions - median 657 KB, mean 52 MB, max 288 MB, 36 % over 10 MB - because 37 % of modern 10-Ds carry a Reg AB II asset-level `EX-102` XML rider. This dataset's bronze is instead SEC's own `-index-headers.html` submission-index object (2.5–6.6 KB, measured), which carries the complete SGML header AND the document manifest and nothing else. That is what keeps the cheap cover-page table from being hostage to the multi-terabyte asset-level one. # Two fields the trust's own block states are deliberately NOT carried A registrant's `STATE OF INCORPORATION` and `EIN` are per-block fields, and the shared header abstraction can scope them to the FIRST block only (`edgar_docs::party_field`). Here the first block is the DEPOSITOR on 19 % of filings, so reading them would record the depositor's domicile as the trust's - and reading them correctly would mean copying `HeaderField`'s spelling table into this collector, which is the failure CLAUDE.md's "fix the abstraction, not the eightieth call site" rule names. For a trust's boilerplate Delaware domicile that trade is not worth making: `issuing_entity_cik` is carried and `edgar_filings` holds both fields for every CIK.

One row per One row per Form 10-D / 10-D/A submission - one distribution period for one asset-backed issuing entity - keyed by its ACCESSION. The key is the accession and NOT the archive path because EDGAR indexes one submission under one path PER associated registrant, and a securitisation associates the trust, the depositor and often a co-issuer: measured on four dissemination days spanning 2015-2026 (master.20150415, 20190815, 20240415, 20260615), 468 index rows carry only 317 distinct submissions. Keying on the path would store 32 % of every period twice. A trust files ~12 rows a year for 3-7 years, so this is a genuine per-deal time series, not a snapshot..

Point-in-time availability

Every Obscura dataset carries available_date: the calendar day the publisher made the row available, day-of, with no session rounding. It is the one column a backtest filters on, and it means the same thing on every dataset in the catalog.

For edgar_abs_distributions: filing_join: available_date is the EDGAR DISSEMINATION day, taken from the bronze object's own day partition, never from the header's FILED AS OF DATE. period_end is CONFORMED PERIOD OF REPORT - SEC's own structured field, and systematically EARLIER than publication (CNH Equipment Trust 2023-A reports the period ending 2024-03-31 in a submission disseminated 2024-04-15), which is exactly why it can never be the availability date. The distribution date stated in the cover page's prose is deliberately NOT parsed, so there is no occurred_on: it is class C by construction and the submission bytes that carry it are 52 MB/filing on average.

Refresh cadence

Obscura refreshes edgar_abs_distributions every 6 hours - the most frequent scheduled job that re-collects or re-exports it. This is Obscura's own pipeline cadence, not the upstream publisher's release schedule; when a row became public is recorded per row in available_date.

Schema - 22 columns

The full public column list for edgar_abs_distributions, with the meaning of every field. The same schema is served unauthenticated at https://api.obscura.trade/v1/catalog/edgar_abs_distributions.

Access edgar_abs_distributions

Two delivery paths, one identifier. Both require an Obscura account and an active subscription; the catalog entry and the schema above are public.

import obscura

client = obscura.Client("obs_live_…")

df = client.query(
    dataset="edgar_abs_distributions",
    symbols=["NVDA", "AAPL"],
    start="2024-01-01",
)

Create a free account Browse all 122 datasets

Frequently asked questions

Related datasets