The Form BW-2 Files Dataset is a closed historical corpus of quarterly periodic reports filed with the SEC by the International Bank for Reconstruction and Development (IBRD, the lending arm of the World Bank) under Section 15 of the Bretton Woods Agreements Act of 1945. Each record represents one EDGAR submission of Form BW-2, keyed by an 18-digit accession number, and packages the full text of the BW-2 quarterly report together with a normalized JSON manifest of the EDGAR header. IBRD is the sole filer; every record is a quarterly report covering its development lending, borrowing programs, liquidity, and financial risk management as of quarter-end. The dataset begins with the earliest BW-2 EDGAR submission on 1996-05-01 and ends with the final filing before the SEC rescinded the form under EDGAR Release 8.6 effective July 2003. Containers are delivered as monthly ZIP archives, and each record folder contains plain-text document files (document-N.txt) and a metadata.json manifest.
Programmatically retrieve the full list of dataset archive files, download URLs and dataset metadata.
Dataset Index JSON API
Download the entire dataset as a single archive file.
Download Entire Dataset:
Download a single container file (e.g. monthly archive) from the dataset.
Download Single Container:
The dataset captures the entire EDGAR life of Form BW-2, a stand-alone periodic report disclosure form created specifically for World Bank debt securities registered under Schedule B of the Securities Act of 1933. Schedule B governs sovereign and supranational issuers, who are exempt from the standard registration regime applicable to private issuers. Unlike Form 10-Q, which interleaves narrative items with prescribed financial statements under Regulation S-X, BW-2 was instructionally lean: it required IBRD to publish, on a quarterly basis, management's discussion and analysis covering its lending and borrowing activities, condensed financial statements, and supplementary disclosures on the development loan portfolio, liquidity, funding resources, and financial risk management. The filing functioned as the World Bank's quarterly investor information document for its debt programs, structured around treasury and credit-risk topics rather than the operating-segment revenue narrative of a commercial 10-Q.
The original SGML submission shipped to EDGAR consisted of a single <DOCUMENT> element of <TYPE>BW-2, containing the entire body as plain text with embedded SGML/HTML-style table markers. There were no separately filed exhibits, no XBRL instance documents, and no structured data files; BW-2 predated the XBRL mandate and was rescinded before structured data became applicable to it. The dataset is delivered as monthly ZIP containers; the file types found inside the records are TXT and JSON. Coverage runs from May 1996 through July 2003, when the SEC rescinded the form under EDGAR Release 8.6, making this a closed, finite corpus with no post-2003 successor on EDGAR.
A single record in the Form BW-2 Files Dataset is one EDGAR submission of Form BW-2 by IBRD, keyed by an 18-digit SEC accession number. Each record packages, as a self-contained folder, the full text of the BW-2 quarterly periodic report as it was disseminated through EDGAR plus a normalized JSON manifest carrying the EDGAR header metadata. Every record is a quarterly report, every record was filed by IBRD (CIK 0000051120, Schedule B file number 083-00003), and every record was generated under the BW-2 disclosure regime tied to Section 15 of the Bretton Woods Agreements Act of 1945.
The dataset is delivered as monthly ZIP archives named YYYY-MM.zip. The ZIP root holds a single directory YYYY-MM/, beneath which each accession is materialized as a folder named with the 18-digit accession number with dashes stripped (for example 000091205700024988 for accession 0000912057-00-024988). Because BW-2 is quarterly and the issuer is a single supranational entity, the great majority of months across the 1996–2003 span have zero filings, and the index only surfaces months where at least one filing exists.
Inside one record folder there are exactly two files:
metadata.json — the normalized EDGAR header, always present.document-N.txt — the body of each <DOCUMENT> in the original SGML submission, numbered sequentially starting at 1. For BW-2, virtually every filing carries a single document, so only document-1.txt is present.Image files referenced in the SGML body as [GRAPHIC OMITTED] are excluded from the dataset; their absence is signalled inline in the TXT body where the original logo or chart appeared.
metadata.json manifestmetadata.json is a flat JSON object that re-expresses the EDGAR submission header as typed fields. It carries the form-level identifiers, filing timestamps, period-of-report date, links back to the original SEC archive, and a per-document inventory of the SGML submission. The principal fields are:
formType — always the literal string "BW-2".accessionNo — the dashed accession number (mirrors the bare-digit folder name).filedAt — ISO 8601 timestamp of acceptance by EDGAR.periodOfReport — the quarter-end the report covers. Because IBRD's fiscal year ends June 30, period dates fall on September 30 (fiscal Q1), December 31 (fiscal Q2), March 31 (fiscal Q3), and June 30 (fiscal year-end).documentFormatFiles — array describing each EDGAR document. The first entry is the actual Form BW-2 document (type: "BW-2", sequence: "1", description: "FORM BW-2", with size matching the byte length of document-1.txt); a trailing entry with blank sequence and type describes the EDGAR "Complete submission text file" wrapper, referenced by URL but not stored inside the record folder.entities — array, almost always a single object describing IBRD as Filer. The stable identifying fields are constant across the dataset: cik 0000051120, companyName INTERNATIONAL BANK FOR RECONSTRUCTION & DEVELOPMENT (Filer), fileNo 083-00003 (the Schedule B file number), sic 6022 State Commercial Banks, act 34 (Securities Exchange Act), and fiscalYearEnd 0630. Per-filing variability is essentially limited to the EDGAR filmNo.linkToTxt, linkToHtml, linkToFilingDetails — SEC archive URLs to the original submission text, the EDGAR index page, and the filing-detail directory.linkToXbrl — an empty string for every record; BW-2 has no XBRL counterpart.dataFiles — an empty array; no structured data files are attached to BW-2 filings.id — a 32-character hex hash that serves as the dataset's internal record identifier.description — a short human-readable label of the form (e.g. "Form BW-2 - Periodic World Bank filing, submitted quarterly").A representative manifest, lightly trimmed, looks like:
1
{
2
"formType": "BW-2",
3
"accessionNo": "0000912057-00-024988",
4
"filedAt": "2000-05-16T00:00:00-04:00",
5
"periodOfReport": "2000-03-31",
6
"documentFormatFiles": [
7
{ "sequence": "1", "size": "64709", "description": "FORM BW-2", "type": "BW-2" },
8
{ "sequence": " ", "size": "65966", "description": "Complete submission text file", "type": " " }
9
],
10
"entities": [
11
{
12
"cik": "0000051120",
13
"companyName": "INTERNATIONAL BANK FOR RECONSTRUCTION & DEVELOPMENT (Filer)",
14
"fileNo": "083-00003",
15
"sic": "6022 State Commercial Banks",
16
"act": "34",
17
"fiscalYearEnd": "0630",
18
"type": "BW-2",
19
"filmNo": "636856"
20
}
21
],
22
"linkToXbrl": "",
23
"dataFiles": []
24
}
document-1.txt bodyThe TXT body is the plain-text inner content of the EDGAR <DOCUMENT> element, with the outer SGML wrappers (<DOCUMENT>, <TYPE>, <SEQUENCE>, <TEXT> … </TEXT> … </DOCUMENT>) stripped. The wrapper-level metadata is surfaced instead through documentFormatFiles[0] in the JSON manifest, so any tooling expecting raw EDGAR <DOCUMENT> envelopes must read sequence, type, size, and description from the manifest rather than from the TXT.
Three non-content markup conventions inherited from the original EDGAR rendering are preserved inline:
<PAGE> markers separating the logical pages of the underlying paginated document.<TABLE> / <CAPTION> / <S> / <C> / </TABLE> markers delimiting financial tables. Inside these markers, columns are visually aligned with whitespace; <S> typically tags the stub (label) column and <C> each numeric value column.[GRAPHIC OMITTED] placeholders standing in for omitted images.Body text is otherwise monospaced ASCII (with occasional 8-bit characters in entity-substituted forms), section headings rendered in uppercase and centered or indented for visual emphasis rather than as semantic markup, and currency figures aligned by whitespace rather than tab-delimited.
The body of a BW-2 filing is structurally stable across the 1996–2003 corpus. A typical record proceeds, in order, through the following blocks:
Cover page. A centered identifying header naming the International Bank for Reconstruction and Development, the document title ("MANAGEMENT'S DISCUSSION & ANALYSIS AND CONDENSED QUARTERLY FINANCIAL STATEMENTS"), the quarter-end date (e.g. "MARCH 31, 2000"), and an "(UNAUDITED)" qualifier.
Table of contents. A plain-text contents listing with page numbers, mirroring the printed quarterly investor report.
Management's Discussion and Analysis (MD&A). A narrative section organized around results of operations and financial highlights. Recurring sub-blocks discuss period-over-period net income, equity, the development loan portfolio, new loan and hedging products, liquidity and funding strategy, and period-specific topics — for example, FY 2000 filings address Year 2000 (Y2K) systems remediation, while later filings discuss changes in IBRD's risk-management framework and treasury-product mix.
IBRD Condensed Financial Statements. Four prescribed statements, each rendered as a <TABLE> block with comparative columns (current quarter-end versus prior fiscal year-end):
Notes to the Financial Statements. Discrete numbered notes elaborating accounting policy and balance-sheet composition. Recurring notes cover the loan portfolio (loans outstanding, undisbursed balances, accumulated provision for loan losses), nonaccruing loans by country (with named obligors such as Bosnia and Herzegovina or Sudan), guarantees (called and uncalled portions), the Fifth Dimension Program supporting IDA-supplementary credits, derivative and hedging instruments, and segment reporting.
Review Report of Independent Accountants. A short signed accountants' review report — not a full audit opinion, since the statements are condensed and unaudited — attesting that the reviewers are not aware of material modifications required for conformity with IBRD's accounting policies.
TALRS appendix tables. A supplementary set of "Treasury Asset Liability Risk System" SEC reports listing detailed bond-level changes in borrowings for the quarter. Each row is a single bond event with columns for BORROWING TYPE, TRADE ID, CURRENCY, EXTERNAL ID, CURRENCY AMOUNT, US$ EQUIVALENT, SETTLEMENT DATE, and MATURITY DATE. The tables are broken out by currency (USD, ITL, JPY, and other denominations) and by lifecycle phase (early retirement vs. new issuance), with currency subtotals and a grand total at the foot.
This ordering is consistent across the corpus; what varies between quarters is the period-of-report dates, the dollar magnitudes, the named topics in MD&A, and the bond-level rows in the TALRS tables.
Cover header:
1
<PAGE>
2
INTERNATIONAL BANK FOR RECONSTRUCTION AND
3
DEVELOPMENT
4
5
MANAGEMENT'S DISCUSSION & ANALYSIS
6
AND
7
CONDENSED QUARTERLY FINANCIAL STATEMENTS
8
MARCH 31, 2000
9
10
(UNAUDITED)
Balance-sheet table fragment showing the <TABLE>/<S>/<C> markup:
1
<TABLE>
2
<CAPTION>
3
MARCH 31,2000 JUNE 30, 1999
4
(UNAUDITED)
5
------------- -------------
6
<S> <C> <C>
7
ASSETS
8
Due from banks $ 734 $ 697
9
Investments--Trading 27,405 30,345
TALRS appendix fragment showing per-bond settlement records:
1
BOND/SELL USD/IBRD/0200USD00.00 0000000488 USD USD0189MLT45 18,000,000 18,000,000 06-Mar-1985 15-Feb-2000
2
BOND/SELL USD/IBRD/0300USD00.00 0000000747 USD USD1384COL01 125,000 125,000 11-Jan-1990 15-Mar-2000
3
-------------
4
TOTAL BY CURRENCY 18,245,000
5
=============
Each record includes the full normalized metadata for the EDGAR submission and the full text of the BW-2 document body as filed. The EDGAR "Complete submission text file" — the SGML envelope wrapping all documents in the submission — is referenced by URL in documentFormatFiles but is not stored inside the record folder; the per-document body in document-1.txt carries the substantive content. Image attachments referenced in the body are not retained, and their positions are marked by inline [GRAPHIC OMITTED] tokens. There are no exhibits to include or exclude — BW-2 does not contemplate an exhibit list — and there is no XBRL or other structured data layer at any point in the form's history.
The disclosure shape of Form BW-2 was institutionally narrow: a single-issuer, quarterly periodic report tied to a discrete statutory regime, and the SEC did not periodically rewrite its instructions in the way it did for Form 10-K and 10-Q. The substantive evolution visible in the dataset is therefore driven less by SEC rule changes and more by IBRD's own evolving accounting and treasury practice between 1996 and 2003. Notable accretions and shifts include the addition of a Statement of Comprehensive Income following adoption of comprehensive-income reporting in IBRD's policies in the late 1990s, the introduction and later expansion of segment-reporting notes, the emergence and disappearance of the Year 2000 remediation discussion in MD&A around fiscal 1999 and 2000, and progressively richer disclosure of derivative and hedging products as IBRD's treasury programs grew. Structurally, the cover-page-through-TALRS sequence remained stable. The form itself was rescinded under EDGAR Release 8.6 effective July 2003, and the dataset closes with the final pre-rescission filing — there are no post-2003 records.
Form BW-2 lived entirely within EDGAR's pre-XBRL, ASCII-SGML era. From the first 1996 filings through the final 2003 filing, the document body is plain ASCII text wrapped in legacy EDGAR SGML markers — <PAGE>, <TABLE>, <CAPTION>, <S>, <C> — with monospaced column alignment for financial tables. The form was rescinded before EDGAR's broader migration to HTML as a primary submission medium became routine for periodic filings, and well before the 2009 inline XBRL mandate; consequently, no record in the dataset carries HTML primary content, no record carries an XBRL instance, and linkToXbrl is uniformly empty. The only format evolution within the corpus is incidental: minor variations in EDGAR's text rendering of tables, occasional drift in the use of [GRAPHIC OMITTED] placeholders, and small differences in whitespace conventions across submission years.
Several characteristics matter for downstream extraction:
entities block in metadata.json is essentially deduplicated; the meaningful per-record metadata variation lives in accessionNo, filedAt, periodOfReport, the documentFormatFiles size and URL fields, and filmNo.periodOfReport must be interpreted against IBRD's June 30 fiscal year-end — a March 31 period-of-report corresponds to IBRD's fiscal Q3, not a calendar Q1.<TABLE> blocks, and stub-column versus value-column distinctions follow the <S> / <C> convention rather than explicit cell delimiters. Negative numbers are typically rendered in parentheses, and currency figures may carry a leading $ only on the first row of a section.dd-Mon-yyyy date strings — and is best treated as a separate parsing target with its own column model.<DOCUMENT> wrapper in the TXT. The absence of an SGML <DOCUMENT> wrapper inside document-1.txt means tools designed to consume raw EDGAR submissions must read sequence, type, size, and description from metadata.json rather than from the body text itself.Every record in this dataset is filed by a single issuer: the International Bank for Reconstruction and Development (IBRD), the lending arm of the World Bank Group. Form BW-2 is not a generally available form. It exists solely to carry IBRD's periodic financial disclosure to U.S. investors in its debt securities. No other World Bank Group entity (IFC, IDA, MIGA) and no other supranational issuer files on it. The filer population is exactly one entity, and each accession number is one quarterly report submitted by IBRD.
The filing obligation rests on a framework specific to IBRD as a supranational debt issuer in the United States.
Form BW-2 is calendar-driven, not event-driven. The trigger is the continuing fact that IBRD has Schedule B-registered debt outstanding and publicly held in U.S. markets. Once that condition holds, IBRD owes a BW-2 covering each interim fiscal quarter, submitted after quarter-end. The form is not produced by transactions, ownership thresholds, or material events; it is produced by the calendar.
Each report typically contains MD&A on financial condition and results of operations, interim financial statements covering lending and borrowing activity, and disclosures on the development loan portfolio, liquidity, funding, and financial risk management as of quarter-end.
The dataset spans the entire EDGAR life of Form BW-2.
The combination of a single filer, a fixed quarterly cadence, and a closed May 1996 to July 2003 window means every record in this dataset is an IBRD quarterly report; there are no insider, third-party, or event-driven submissions in this corpus.
Form BW-2 occupies a narrow corner of EDGAR. The most informative comparisons are to its annual companion (BW-3), the periodic forms that BW-2 functionally resembles for other issuer classes (10-Q, 20-F/6-K, 18-K), the registration regime behind the underlying securities (Schedule B vs. Schedule A), and other specialty disclosure tracks (ABS-EE, ABS-15G) that mark the outer boundary of comparable regimes.
Same issuer, same statutory basis, same rescission (EDGAR Release 8.6, July 2003). BW-3 carries audited annual financials and broader narrative; BW-2 carries quarterly interim financials and updates on lending, borrowing, liquidity, and risk. Complementary, not overlapping: BW-3 cannot stand in for BW-2's interim coverage.
The functional analog for U.S. operating companies under Exchange Act Section 13/15(d). Both are quarterly with interim financials and MD&A, but the differences are foundational:
10-Q and BW-2 are not substitutable in either direction.
The natural comparison for non-U.S. SEC registrants. BW-2 differs on three axes: the filer is a multilateral institution, not a foreign private issuer; authority is the Bretton Woods Agreements Act, not Section 13(a) or 15(d); and 6-K is event-driven, not scheduled, while BW-2 is strictly quarterly periodic. 20-F annual filings carry home-country reconciliation requirements that have no World Bank analog. World Bank disclosure does not appear in 20-F or 6-K datasets at all.
The closest sovereign-style analogs, covering foreign governments and political subdivisions issuing registered debt. Shared territory: non-corporate issuer, debt-service and fiscal disclosure rooted in Schedule B. Key differences:
Schedule A of the Securities Act governs ordinary corporate registration (S-1, S-3, etc.); Schedule B governs foreign government issuers and, by statutory extension, World Bank debt. BW-2 is not a registration statement but the ongoing periodic disclosure attached to Schedule B-registered securities. The comparison explains why BW-2 line items follow sovereign- and lender-style conventions rather than the operating-company conventions of 10-K/10-Q.
Useful only as boundary markers. Like BW-2, they sit outside the 10-K/10-Q/8-K framework and serve a narrow filer population. They differ on every substantive axis: ABS-EE covers asset-level data for securitized pools under Regulation AB II in structured XML; ABS-15G reports ABS repurchase activity under Exchange Act Section 15G. Both are active regimes; BW-2 is rescinded. They are not analytical substitutes.
BW-2 is the only quarterly periodic-report family on EDGAR filed by IBRD under the Bretton Woods Agreements Act, covering development lending, borrowing programs, and liquidity rather than commercial operations. The dataset is closed in time (May 1996 to July 2003) and has no EDGAR-resident successor of equivalent scope. BW-3 is its only true companion; 10-Q, 20-F/6-K, and 18-K are functional analogs for other issuer classes but are not interchangeable; Schedule B explains the registration lineage; ABS-EE/ABS-15G show that specialty regimes exist but address unrelated subject matter. Treat BW-2 as a self-contained historical archive of World Bank quarterly disclosure with no current-day SEC substitute.
A closed corpus of IBRD quarterly filings spanning May 1996 to July 2003 attracts a narrow but identifiable set of professional users. Each anchors on a different slice of the filings.
Sell-side and buy-side analysts covering supranational, sovereign, and sub-sovereign debt use the filings as a longitudinal credit record. They pull from the MD&A, lending and borrowing schedules, loan-loss provisioning, and liquidity disclosures to reconstruct historical capital-to-loans ratios, callable capital trends, and liquid-asset coverage, and to build long-run series that pre-date current investor materials and bridge into later Form 18-K disclosures.
Treasury and ALM desks holding multilateral paper as HQLA use the financial statements, derivatives exposure tables, and currency/rate risk narratives to support internal credit reviews, counterparty-limit files, and the evidence trail for zero-risk-weight or HQLA-1 treatment under regulatory capital and liquidity reporting.
Researchers studying multilateral lending use the development loan portfolio disclosures, country exposure summaries, disbursement schedules, and operational narrative as primary sources. Typical outputs are panel datasets linking quarterly lending to borrower-country macro outcomes, studies of issuer behavior during the late-1990s emerging-market crises, and historical analyses of multilateral capital adequacy frameworks.
Scholars and practitioners tracking derivatives disclosure across the late-1990s accounting transition use the financial risk management narrative, swap notional and fair-value tables, and accounting policy notes to study how a sophisticated supranational described swap programs and hedge accounting before and after fair-value standards took effect.
The 1998-2000 filings contain Y2K remediation, technology risk, and contingency funding narrative. Researchers comparing Y2K disclosure across issuer types use these as a supranational reference point that is otherwise hard to reconstruct.
Researchers documenting EDGAR's evolution and the regulatory treatment of non-standard issuers use the dataset as a complete record of a rescinded form. They work from the metadata JSON (filing dates, accession numbers, submission headers) and the timing around EDGAR Release 8.6 to map form lifecycles and issuer-specific disclosure regimes onto the broader filing taxonomy.
Legal researchers studying the Schedule B / Bretton Woods Agreements Act regime use the corpus to examine mandated disclosure form and content, exemptions from standard registration, and the interaction between Section 15 of the 1945 Act and the 1933 Securities Act. Outputs include law-review articles and advisory materials for other supranational or sovereign issuers contemplating U.S. registered offerings.
Methodology and policy staff at rating agencies use the filings to back-test supranational rating criteria, focusing on capital adequacy ratios, callable capital structure, preferred-creditor-treatment narrative, and arrears disclosures, particularly across periods of concentrated crisis-borrower exposure.
Teams building reference corpora for financial language models and document-classification pipelines treat the dataset as a finite, single-issuer example of a rare, rescinded form type. They use the metadata JSON, raw TXT submission structure, and consistent section headings to build gold-standard labels, train form-type classifiers that must route legacy filings, and assemble retrieval indices with full EDGAR coverage.
Users split cleanly across financial statements (analysts, treasury, rating teams), narrative disclosures (academics, legal and Y2K researchers), and metadata and structure (EDGAR historians, NLP curators). Each group draws on a distinct cross-section to support credit memos, ALM documentation, academic papers, legal commentary, or labeled training data.
The Form BW-2 Files Dataset supports a small set of concrete, repeatable workflows tied to specific sections of each BW-2 record.
Building a longitudinal IBRD capital-adequacy series. Pull Loans outstanding, accumulated provision for loan losses, equity, and callable-capital references from the Balance Sheet <TABLE> blocks and MD&A across all quarters in the corpus to construct a 1996-2003 capital-to-loans and equity-to-loans time series. The output feeds SSA credit memos, rating-agency back-tests of supranational criteria, and the pre-2003 leg of long-run charts that splice into later Form 18-K data.
Quarter-by-quarter bond issuance and early-retirement reconstruction. Parse the TALRS appendix tables, splitting the wide fixed-width rows on BORROWING TYPE, TRADE ID, CURRENCY, EXTERNAL ID, US$ EQUIVALENT, SETTLEMENT DATE, and MATURITY DATE. The result is a bond-event-level panel of new issuance versus early retirement by currency (USD, JPY, ITL, others) used by treasury researchers and DCM strategists studying World Bank funding mix and tactical buyback behavior across the late-1990s rate cycle.
Crisis-window exposure tracking via the nonaccruing-loans note. Extract the named-obligor lists from the nonaccruing-loans note (e.g., Bosnia and Herzegovina, Sudan) and the development-loan-portfolio MD&A across the 1997-1999 emerging-markets-crisis quarters. This supports academic panels linking IBRD's portfolio response to borrower-country macro events and feeds preferred-creditor-treatment commentary in rating reviews.
Derivatives and hedge-accounting transition study. Mine the financial risk management narrative and the derivative/hedging notes for swap notionals, fair-value disclosure language, and accounting-policy text spanning the SFAS 133 adoption window. The corpus gives derivatives researchers a single sophisticated supranational issuer's before/after disclosure under stable form instructions.
Y2K supranational reference point. Filter records by periodOfReport to fiscal 1999 and 2000 quarters and extract the Y2K remediation and contingency-funding sub-blocks from MD&A. These passages anchor cross-issuer Y2K disclosure studies that otherwise lack a multilateral lender comparator.
Form-lifecycle and EDGAR-history mapping. Use metadata.json fields (filedAt, accessionNo, filmNo, documentFormatFiles, empty linkToXbrl and dataFiles) across the corpus to chart the BW-2 cadence from May 1996 through the July 2003 rescission under EDGAR Release 8.6. The output supports SEC-form historians documenting rescinded-form lifecycles and NLP teams labeling legacy form types in retrieval indices.
Gold-standard training data for legacy-form classifiers. Treat the closed BW-2 record set as a labeled example bucket for a rare form type. The plain-ASCII body, stable cover-page-through-TALRS section ordering, and <PAGE>/<TABLE>/<S>/<C> markup conventions give corpus curators a clean, finite training and evaluation slice for form-type routing and section-segmentation models on pre-XBRL EDGAR submissions.
The Form BW-2 Files Dataset can be accessed in three ways: through the dataset index JSON API to retrieve metadata and the list of available containers, by downloading the full dataset archive, or by downloading individual container files. Because the historical corpus is small, the entire archive is lightweight and easy to retrieve in a single request.
Dataset Index JSON API: https://api.sec-api.io/datasets/form-bw2-files.json
This endpoint returns dataset metadata including the dataset name, description, last updated timestamp, earliest sample date, total record and size counters, form types covered, container format, and file types. It also returns the download URL for the full archive along with the complete list of individual container files. Each container entry includes its key, size, record count, updated timestamp, and direct download URL. This endpoint does not require an API key. It can be polled regularly to detect which containers were modified in the latest refresh, allowing incremental downloads on a per-container basis.
1
{
2
"datasetId": "1f13365b-9ae0-6a53-86fe-8c4a8f2d84c4",
3
"datasetDownloadUrl": "https://api.sec-api.io/datasets/form-bw2-files.zip",
4
"name": "Form BW-2 Files Dataset",
5
"updatedAt": "2026-04-16T08:45:21.281Z",
6
"earliestSampleDate": "1996-05-01",
7
"totalRecords": 27,
8
"totalSize": 876786,
9
"formTypes": ["BW-2"],
10
"containerFormat": "ZIP",
11
"fileTypes": ["TXT", "JSON"],
12
"containers": [
13
{
14
"downloadUrl": "https://api.sec-api.io/datasets/form-bw2-files/2003/2003-07.zip",
15
"key": "2003/2003-07.zip",
16
"size": 32145,
17
"records": 1,
18
"updatedAt": "2026-04-16T08:45:21.281Z"
19
}
20
]
21
}
Download Entire Dataset: https://api.sec-api.io/datasets/form-bw2-files.zip?token=YOUR_API_KEY
This endpoint returns the complete dataset as a single ZIP archive containing all monthly containers from May 1996 through July 2003. The full archive is small enough to download in one request and store locally without needing incremental retrieval. This endpoint requires an API key.
Download Single Container: https://api.sec-api.io/datasets/form-bw2-files/2003/2003-07.zip?token=YOUR_API_KEY
Each container is a monthly ZIP file containing all Form BW-2 filings submitted to EDGAR during that month, including a metadata file and the original submission documents (excluding image files) for each accession number. Use the container key values returned by the index JSON API to construct download URLs for any specific month. This endpoint requires an API key.
The Form BW-2 Files Dataset covers Form BW-2, the quarterly periodic report filed with the SEC by the International Bank for Reconstruction and Development (IBRD, the World Bank) under Section 15 of the Bretton Woods Agreements Act of 1945. It is the only form type in the dataset, and it was rescinded by the SEC under EDGAR Release 8.6 effective July 2003.
A single record is one EDGAR submission of Form BW-2 by IBRD, keyed by an 18-digit SEC accession number. Each record is a self-contained folder containing a metadata.json manifest of the EDGAR header and a document-1.txt file holding the plain-text body of the BW-2 quarterly report.
Only the International Bank for Reconstruction and Development (IBRD), CIK 0000051120, Schedule B file number 083-00003. Form BW-2 is not a generally available form; no other World Bank Group entity (IFC, IDA, MIGA) and no other supranational issuer files on it. The filer population is exactly one entity.
The dataset spans the entire EDGAR life of Form BW-2: from the earliest sample date of 1996-05-01 through the final pre-rescission filing in July 2003, when the SEC rescinded the form under EDGAR Release 8.6. There are no post-2003 records.
The dataset is delivered as monthly ZIP containers. Each record folder contains two file types: a metadata.json manifest that re-expresses the EDGAR submission header as typed JSON fields, and one or more document-N.txt files holding the plain-ASCII body of the BW-2 submission with legacy SGML markers such as <PAGE>, <TABLE>, <S>, and <C> preserved inline. There is no XBRL or other structured data layer; linkToXbrl is uniformly empty.
Form 10-Q is the quarterly periodic report filed by U.S. operating companies under Exchange Act Section 13/15(d), built on Regulation S-X / S-K and tagged with XBRL. Form BW-2 was filed by a single supranational lender under a 1945 statutory carve-out, structured around development-lending and borrowing-portfolio disclosure rather than commercial-operations narrative, and submitted as plain text with no XBRL tagging. The two forms are not substitutable in either direction.
Form BW-3 is the annual companion to BW-2. Both share the same filer (IBRD), the same statutory basis (Section 15 of the Bretton Woods Agreements Act of 1945), and the same rescission date (EDGAR Release 8.6, July 2003). BW-3 carries audited annual financials and broader narrative; BW-2 carries quarterly interim financials and updates on lending, borrowing, liquidity, and risk. They are complementary, not overlapping, and BW-3 filings are not part of this dataset.