Form 13F-E Files Dataset

The Form 13F-E Files Dataset is a closed, EDGAR-native archive of the quarterly institutional-holdings reports that institutional investment managers filed electronically under Section 13(f) of the Securities Exchange Act of 1934 before mandatory electronic filing took effect. Each record corresponds to a single EDGAR submission — either an original 13F-E holdings report or a 13F-E/A amendment — and preserves the manager's cover page, signature block, other-managers list, and the fixed-column information table of Section 13(f) securities held as of the quarter-end reporting date. The dataset covers the entire life of the 13F-E submission type, from the start of EDGAR voluntary filing in January 1994 through its retirement in April 1999, when the SEC replaced the single 13F-E suffix with the modern 13F-HR (Holdings Report) and 13F-NT (Notice) scheme. Filings are distributed as ZIP containers of TXT and JSON files and form the only EDGAR-native source for pre-April-1999 manager-level institutional holdings.

Dataset APIs

Dataset Files

What This Dataset Contains

The dataset captures Form 13F-E, the EDGAR submission type used by institutional investment managers to satisfy the quarterly public-disclosure requirement of Section 13(f) of the Securities Exchange Act of 1934. It applied to managers exercising investment discretion over accounts holding $100 million or more in Section 13(f) securities — primarily U.S. exchange-listed equities plus certain convertible debt, exchange-traded options, and warrants enumerated on the Commission's official 13(f) securities list. Form 13F-E was the pre-mandatory-electronic-filing variant used from the start of EDGAR voluntary filing in 1994 through April 1999, when the Commission replaced it with the current two-variant scheme — Form 13F-HR (Holdings Report) and Form 13F-NT (Notice when holdings are reported by another manager). The /A suffix in 13F-E/A denotes an amendment that restates, corrects, or supplements a previously filed 13F-E; amendments carry an explicit AMENDMENT NN annotation on the cover page but otherwise reproduce the same anatomy as an original.

Each filing comprises two tightly linked logical parts: a narrative cover page that identifies the reporting manager, enumerates any co-filers, and carries the signed certification, and a structured information table that lists every Section 13(f) security held as of the quarter-end reporting date with share or principal quantities, market values, investment-discretion codes, manager-index cross-references, and voting-authority breakdowns. All documents in the dataset are fixed-column ASCII plaintext — Form 13F-E never used HTML, never used XBRL, and predates the XML information-table schema introduced years later for 13F-HR. Filings are delivered as ZIP containers holding TXT bodies alongside JSON headers.

Content Structure of a Single Record

What one record represents

One record in the Form 13F-E Files dataset corresponds to a single EDGAR submission identified by one accession number — either an original Form 13F-E quarterly institutional-holdings report or a Form 13F-E/A amendment to a previously filed 13F-E. On disk, a record is a folder whose name is the 18-digit zero-padded accession number with dashes stripped (accession 0000068100-98-000375 becomes folder 000006810098000375). Every folder contains exactly one metadata.json header file plus one or more document-<sequence>.txt files, where <sequence> matches the <SEQUENCE> integer assigned to each document by the original EDGAR SGML submission. The textual bodies of the submission are retained; graphic attachments (images that were wrapped in <TYPE>GRAPHIC blocks in the SGML) are not materialized.

Record-level layout

At the record level, content is organized in two concrete layers.

1. metadata.json — a structured JSON header derived from the EDGAR SGML submission wrapper. It carries filing identity, filing and reporting dates, EDGAR archive links, a documentFormatFiles manifest, and an entities array describing the filer and any co-filers.
2. document-<sequence>.txt — the stripped filing body or bodies. The outer SGML <SUBMISSION>, <DOCUMENT>, <TYPE>, and <TEXT> envelopes have already been removed, so each .txt file begins directly with the document body. The primary document-1.txt is the Form 13F-E itself, containing the cover page and the holdings information table in the legacy fixed-column plaintext format used on EDGAR before the HTML and XBRL eras, lightly sprinkled with <PAGE> pagination markers and <TABLE>…</TABLE> tags that wrap the tabular region.

metadata.json

The header groups fields into the following clusters.

• Filing identity

  • formType — 13F-E for originals or 13F-E/A for amendments.
  • accessionNo — canonical dashed accession number (e.g., 0000068100-98-000375).
  • id — 32-character internal hash identifier.
  • filedAt — ISO-8601 timestamp with Eastern-time offset (e.g., 1998-12-10T00:00:00-05:00).
  • periodOfReport — quarter-end date the report covers (e.g., 1998-09-30).
  • description — human-readable form description, e.g., "Form 13F-E - Quarterly reports filed by institutional managers."

• Links back to EDGAR

  • linkToFilingDetails — the sec.gov archive folder URL.
  • linkToTxt — the complete submission .txt bundle on EDGAR.
  • linkToHtml — the EDGAR index page for the filing.
  • linkToXbrl — an empty string for every record, since Form 13F-E predates XBRL.

• documentFormatFiles — an array describing each document in the submission. Each entry carries a sequence string (matching the document-<sequence>.txt filename on disk), a stringified byte size, a documentUrl pointing to the EDGAR-hosted copy, a type, and an optional description. A trailing entry whose sequence is a single space (" ") references the complete submission .txt bundle on EDGAR rather than a physical document; it is a link, not a file that appears in the folder.

• dataFiles — an empty array. No structured financial-data attachments (XBRL instance documents, schemas, or calculation/label/presentation linkbases) accompany 13F-E filings, and none have been generated retroactively.

• seriesAndClassesContractsInformation — an empty array. This slot exists to hold investment-company series/class identifiers on N-series forms and is unused for 13F-E.

• entities — an array of EDGAR header blocks, one per filer or co-filer. Populated fields typically include cik, companyName (with a role suffix such as "(Filer)"), type (repeats the form type), fileNo (SEC file number, e.g., 028-00490), filmNo (EDGAR film number), irsNo, fiscalYearEnd as a four-digit MMDD string, stateOfIncorporation as a two-letter state code, act (34 for the Exchange Act), and optionally sic (SIC code with its industry label) and tickers (array of ticker symbols). SIC codes and ticker arrays appear only when the filer is itself a publicly listed issuer for which EDGAR supplied that metadata (e.g., a large commercial bank carrying ticker JPM and SIC 6022 State Commercial Banks); for private investment partnerships and non-issuer advisers these fields are simply absent. Missing keys should be treated as "not provided" rather than null.

document-1.txt — cover page (Page 1)

The cover page is narrative plaintext. It opens with the SECURITIES AND EXCHANGE COMMISSION / FORM 13F-E banner and contains:

• The reporting-period caption: Report for the Calendar Year or Quarter Ended: <MM/DD/YY>, followed by an AMENDMENT NN label when the filing is a 13F-E/A.
• The Institutional Investment Manager: identification block with the manager's legal name, street address, city, state, and ZIP code.
• A representation paragraph certifying the report, followed by a signature block carrying signatory name, title, telephone number, signed name, place, and date.
• An Other Managers on Whose Behalf this Report is Filed: list enumerating each additional manager with a two-digit index number and, where applicable, that manager's own SEC file number (e.g., 028-01482). These index numbers are the key cross-reference consumed by the -MANAGERS- column in the information table.

document-1.txt — information table (Page 2 onward)

The information table is the core disclosure and is laid out as a fixed-column plaintext grid paginated by <PAGE> markers. Each table page begins with the canonical two-line column header:

          1
          VALUE    SHARES/ SH/ PUT/ INVSTMT            -----VOTING AUTHORITY-----
        

          2
          NAME OF ISSUER                 -TITLE OF CLASS- --CUSIP-- (x$1000)  PRN AMT PRN CALL DSCRETN -MANAGERS-      SOLE   SHARED    NONE
        

Each data row represents one Section 13(f) security position and encodes, at fixed character offsets:

• Name of issuer — the issuer name, roughly 30 characters wide, left-aligned.
• Title of class — the security-class designator (COMMON, COM, CL A, PFD, NT, etc.).
• CUSIP — the 9-character CUSIP identifying the specific security.
• Value (x$1000) — reported market value expressed in thousands of U.S. dollars as an integer.
• Shares/PRN amount together with the SH/PRN unit flag — share count for equities, principal amount for debt instruments.
• PUT/CALL indicator — blank for cash-equity positions; PUT or CALL for option positions.
• INVSTMT DSCRETN — the investment-discretion code (DEFINED, OTHER, SHARED-DEFINED, SHARED-OTHER, etc.).
• -MANAGERS- — one or more two-digit manager indices that cross-reference the cover page's "Other Managers on Whose Behalf this Report is Filed" list.
• Voting authority — three integer columns (SOLE, SHARED, NONE) that partition the share count across voting-authority categories.

The table closes with a REPORT SUMMARY line that aggregates the entire filing — total number of data records, total aggregate value (in $1000s), and count of other managers on whose behalf the report is filed. A representative summary line is:

          1
          REPORT SUMMARY                  2 DATA RECORDS                6503           15 OTHER MANAGERS ON WHOSE BEHALF REPORT IS FILED
        

A closing </TABLE> tag then terminates the document. The number of rows varies widely: active managers routinely report hundreds of positions, and the same issuer/CUSIP may appear across multiple rows when holdings are split across managers, discretion categories, or voting-authority groupings.

Additional documents (document-2.txt, document-3.txt, …)

When the original submission carried additional textual documents — typically cover letters or narrative exhibits — each is emitted as its own document-<sequence>.txt using the sequence number EDGAR assigned in the SGML submission. These files follow the same plaintext conventions and begin directly with the document body.

Included content

Every record in the dataset includes:

• The complete metadata.json header synthesized from the EDGAR submission wrapper.
• The full Form 13F-E body (cover page, manager identification, signature block, other-managers list, complete information table, and REPORT SUMMARY) delivered as document-1.txt.
• Any additional textual documents that were part of the original submission (cover letters, exhibits), each as its own document-<sequence>.txt retaining the sequence number EDGAR assigned.

Excluded or separate content

• Graphic attachments — images that appeared as <TYPE>GRAPHIC blocks in the original SGML (GIFs, JPEGs) are not materialized on disk.
• The outer SGML envelope — <SUBMISSION>, <DOCUMENT>, <TYPE>, and <TEXT> tags have been stripped so each .txt file begins directly with the document body rather than with SGML wrappers.
• The full-submission .txt bundle — the documentFormatFiles entry whose sequence is " " references EDGAR's combined submission blob and is not materialized as a physical file in the folder.
• XBRL instance documents and linkbases — none exist for this form type. linkToXbrl is always empty and dataFiles is always an empty array.
• Series/class identifiers — seriesAndClassesContractsInformation is always empty for 13F-E (the field is reserved for N-series investment-company forms).

Structural evolution within the 13F-E era

Form 13F-E was stable during its EDGAR lifetime (January 1994 through April 1999). The column layout of the information table — NAME OF ISSUER, TITLE OF CLASS, CUSIP, VALUE (x$1000), SHARES/PRN AMT, SH/PRN, PUT/CALL, INVSTMT DSCRETN, -MANAGERS-, SOLE, SHARED, NONE — remained the canonical disclosure grid across the period and carried forward in substance into the successor Form 13F-HR paper layout. The principal structural distinction within the dataset is between originals (13F-E) and amendments (13F-E/A); amendments add an AMENDMENT NN annotation on the cover page but otherwise share the same anatomy. In April 1999 the form was discontinued and replaced by 13F-HR and 13F-NT under the revised Regulation 13F framework; the XML/structured information-table schema that the Commission later layered onto 13F-HR is not part of this dataset, which closes at the April 1999 discontinuation boundary.

Data format

All documents in this dataset are fixed-column ASCII plaintext. Form 13F-E never used HTML, never used XBRL, and predates the XML information-table schema introduced years later for 13F-HR. The only markup inside document-1.txt files consists of legacy SGML pagination tags — <PAGE> for page breaks and <TABLE> / </TABLE> wrapping the tabular region — inherited from EDGAR's pre-HTML dialect. Holdings parsing therefore requires a column-offset/fixed-width strategy rather than an XML or HTML parser. Monetary values are integers in thousands of U.S. dollars; share and principal quantities are integers accompanied by a SH or PRN unit flag in a separate column.

Interpretation and extraction notes

• One record = one accession-number folder. Every folder contains metadata.json; the primary filing body is always document-1.txt.
• Amendments reference but do not include the original. A 13F-E/A record stands alone; reconstructing an amendment chain requires matching on manager CIK and periodOfReport across records and using the cover-page AMENDMENT NN label (together with filedAt) to order them.
• documentFormatFiles is authoritative for document-to-file mapping. The sequence field pairs directly with the document-<sequence>.txt filename. The trailing entry whose sequence is a single space is a reference to the full-submission bundle on EDGAR, not a physical file inside the folder.
• Co-filing and manager indexing drive holdings interpretation. The -MANAGERS- column in each row references the two-digit indices defined on the cover page's "Other Managers on Whose Behalf this Report is Filed" list. A single CUSIP may appear in multiple rows when positions are split across managers, discretion categories, or voting-authority groupings; these are not duplicates but distinct reporting slices that must be aggregated with care.
• Voting-authority integers should sum to the share count. In well-formed filings SOLE + SHARED + NONE equals the reported SHARES/PRN AMT. Legacy plaintext filings contain filer-level formatting inconsistencies and should be handled defensively.
• The REPORT SUMMARY line is a useful extraction checksum. Its declared number of data records and aggregate value should reconcile with parsed rows; discrepancies typically indicate a column-offset parsing problem rather than a filer error.
• Entity-field presence is filer-dependent. sic and tickers appear in entities[*] only when the filer is itself a publicly listed issuer for which EDGAR supplied issuer metadata; their absence is expected for the large population of private investment advisers and partnerships that file 13F reports.
• filedAt is EDGAR's receipt timestamp, not the as-of date of the holdings. The holdings-as-of date is carried by periodOfReport and by the cover-page Report for the Calendar Year or Quarter Ended: caption; the two should agree and are independently useful for time-series alignment.
• Row-level dollar values are in thousands. Reconstructing a position's full dollar exposure requires multiplying the VALUE column by 1,000; share and principal amounts are reported in their native units.

Who Files or Publishes This Dataset, and When

Who files

The filer is an institutional investment manager with investment discretion over Section 13(f) securities. Under Section 13(f) of the Securities Exchange Act of 1934 (added by the Securities Acts Amendments of 1975), the category covers any entity that invests in securities for its own account, and any person or entity that exercises investment discretion over another's account.

In this dataset the filer population is effectively:

• Registered investment advisers managing funds, separate accounts, or pension assets.
• Banks and bank holding companies with trust or fiduciary departments.
• Insurance companies managing general- and separate-account assets.
• Broker-dealers with discretion over customer accounts.
• Pension plans, endowments, foundations, and corporations managing their own portfolios.
• Trustees and other fiduciaries with investment discretion.
• Natural persons with discretion over qualifying accounts (rare in practice).

The legal filer is the manager with discretion, not the underlying clients, funds, or the issuers of the reported securities. Where an adviser has discretion over a fund's assets, the adviser files and the fund does not.

What triggers the obligation

The filing obligation attaches when a manager exercises investment discretion over accounts holding, in aggregate, at least $100 million in Section 13(f) securities on the last trading day of any month in a calendar year. Once the threshold is crossed in year Y:

• An initial report is due for the fourth quarter of year Y (period of report December 31).
• Reports are then due for Q1, Q2, and Q3 of year Y+1 (periods ending March 31, June 30, September 30).
• The manager files every calendar quarter thereafter for as long as it remains subject, regardless of intra-year fluctuations below $100 million.

"Section 13(f) securities" is a defined universe published quarterly by the SEC on its Official List. It covers exchange-listed and NASDAQ equities, certain equity options and warrants, closed-end fund shares, and certain convertible debt. Corporate debt, open-end fund shares, and most non-U.S.-listed securities are excluded and do not count toward the threshold.

When each record is filed

Form 13F is a periodic report tied to the calendar quarter. Each report:

• Is due within 45 calendar days after the end of the calendar quarter.
• Reflects holdings as of the last day of that quarter.
• Uses the same deadline for every manager; there is no accelerated or smaller-reporting variant.

Why the "E" suffix exists

Before April 1999, electronic filing of Form 13F was permitted but not required. Managers that chose to file on EDGAR used the 13F-E submission type (the "E" marks an electronic submission), while most 13F reports were still filed on paper. In April 1999 the SEC made electronic 13F filing mandatory and retired the 13F-E submission type, replacing it with:

• 13F-HR (Holdings Report) — the manager's own full holdings report with an information table.
• 13F-NT (Notice) — filed when another manager's 13F-HR will report all of this manager's Section 13(f) holdings.

The 13F-E submission type had no HR/NT split; cover-page indicators handled the distinction between filers reporting their own holdings and filers relying on another manager's report. The 13F-E corpus therefore mixes what are now HR- and NT-style reports under a single submission type. This dataset's temporal bounds follow the life of 13F-E: earliest records from 1994, latest from April 1999.

Form 13F-E/A amendments

13F-E/A is the amendment counterpart to 13F-E. A manager files an amendment to correct or supplement a prior 13F-E, typically to:

• Fix issuer names, CUSIPs, share counts, or market values.
• Add omitted holdings or restate a misclassified security.
• Replace the report after staff comments or internal review.
• Release holdings whose confidential treatment request expired, was denied, or was withdrawn.

An amendment keeps the original period of report and does not create a new reporting period. Each 13F-E/A is a separate EDGAR record with its own accession number.

Important distinctions

• Not Schedules 13D and 13G. Those filings are triggered by beneficial ownership above 5 percent of a class of voting equity and are filed by the beneficial owner. A 13F manager may have no 13D/13G obligation, and a 5 percent owner may not meet the institutional-manager definition.
• Not fund filings. Form N-PORT, N-CSR, and other investment company reports are filed by registered funds themselves. A fund advised by a 13F filer does not file 13F; its holdings are reported through the adviser.
• Clients and issuers do not file. Beneficial owners, underlying funds, and the issuers of reported securities are not 13F filers. The same positions can appear on multiple managers' 13Fs where discretion is shared across advisers and sub-advisers.
• Confidential treatment. Historically some managers obtained confidential treatment for specific holdings (often acquisition programs). Those line items did not appear in the public information table for the quarter and sometimes surfaced later through an amendment.
• Non-U.S. managers and natural persons are in scope if they meet the discretion and threshold tests and use U.S. jurisdictional means, but are rare in this corpus.

How This Dataset Differs From Similar Datasets or Filings

Form 13F-E is a closed-universe legacy form used on EDGAR from 1994 through April 1999, when the SEC retired the single 13F-E suffix and split institutional reporting into 13F-HR (holdings) and 13F-NT (notice). The most useful comparisons are to the other 13F suffixes that replaced it, the 13F amendment form, and the small set of neighboring holdings disclosures that filers often confuse with 13F.

Form 13F-HR (holdings report, post-April 1999)

The direct functional successor. 13F-HR shares with 13F-E the same filer population (institutional investment managers), the same $100 million discretionary-AUM threshold, the same quarterly cadence, the same 45-day filing deadline, and the same Section 13(f) list of eligible securities.

Two differences matter:

• Era. 13F-E is the only EDGAR-native source for 1994 through April 1999; 13F-HR is the only source from Q2 1999 onward. A continuous institutional-holdings panel must bridge the two at that boundary.
• Format. 13F-E filings are unstructured fixed-column TXT documents. 13F-HR has been required in structured XML since 2013, with tagged infoTable elements (issuer, CUSIP, value, shares, put/call, discretion, voting authority). Pre-2013 13F-HR filings are also TXT but follow the post-1999 layout rather than the 13F-E format.

Form 13F-NT (notice filing, post-April 1999)

13F-NT exists only after the April 1999 suffix split. It is filed when another manager reports the positions in full, and it contains no holdings table. Under the pre-1999 regime there was no NT suffix: a single 13F-E filing could serve as either a full holdings report or a notice, with the distinction buried inside the document body. Consequently, 13F-E conflates what is now cleanly split between 13F-HR and 13F-NT, and holdings-vs-notice status must be inferred from the filing text rather than the form code.

Form 13F-HR/A and Form 13F-E/A (amendments)

13F-E/A is the pre-1999 amendment form and is included in this dataset; Form 13F-HR/A is its modern equivalent. Both restate or supplement previously filed holdings. The difference is structural: 13F-HR/A carries tagged amendmentType and amendmentNumber XML fields that support programmatic chaining from original to amendment, while 13F-E/A encodes amendment intent in free-form header or cover text, requiring heuristic linkage to the original 13F-E.

Schedules 13D and 13G (beneficial ownership over 5%)

Often confused with 13F because both disclose equity ownership, but the statutory basis and trigger are different. 13D/13G are filed under Section 13(d)/(g) by any person — institutional or not — crossing 5% beneficial ownership of a registered class, and they are event-driven (Schedule 13D within 10 days; Schedule 13G annual or accelerated by filer category). They describe a single issuer, identify the acquirer and group, and state purpose; 13F-E is a portfolio-wide quarter-end snapshot across many issuers with no issuer-level narrative. A manager can appear in both, but the records are not substitutes.

Form N-Q (2004–2019) and Form N-PORT (2019–)

Both are Investment Company Act filings covering 40-Act fund portfolios, not manager-level discretionary holdings. They cover the entire fund portfolio, including non-13(f) securities (fixed income, derivatives, non-U.S. equities), at a fund rather than manager level. The filer populations overlap only partially: an advisor running a mutual fund files N-Q/N-PORT at the fund level and rolls the same positions into its manager-level 13F. Neither has any temporal overlap with 13F-E — Form N-Q was adopted in 2004 and N-PORT in 2019 — so for the 1994–1999 window, 13F-E has no fund-side EDGAR analog.

Form N-CSR (fund shareholder reports, 2003–)

Form N-CSR is a semiannual certified shareholder report for registered investment companies and contains a schedule of investments alongside audited financials and narrative. Overlap with 13F-E is limited to the presence of a holdings table; N-CSR is fund-level, audited, narrative-heavy, and on a different cadence under a different statute. Adopted in 2003, it has no temporal overlap with 13F-E.

Boundary summary

Form 13F-E is defined by a specific intersection: EDGAR-era institutional holdings, at the manager-portfolio level, across Section 13(f) securities, for the 1994 through April 1999 window, in a pre-XML fixed-column TXT format, with no HR/NT suffix split and free-form amendment references.

It is interchangeable with no other dataset:

• 13F-HR and 13F-NT cover the same regime but only post-April 1999, and in structured form from 2013.
• 13D/13G answer a different question (concentrated stakes at a single issuer) under a different statute.
• N-Q, N-PORT, and N-CSR operate under the Investment Company Act, cover fund rather than manager portfolios, include non-13(f) securities, and post-date the 13F-E window entirely.

Any longitudinal institutional-holdings analysis that extends before April 1999 must use 13F-E; any such analysis that extends after it must bridge to 13F-HR and 13F-NT at that boundary.

Who Uses This Dataset

The Form 13F-E Files Dataset is the historical anchor for any 13F time series that extends into the modern 13F-HR dataset regime, and it serves a narrow set of professionals whose work genuinely requires pre-1999 primary-source holdings data.

Quantitative researchers building long-horizon backtests

Quants developing factor models, ownership-based signals, and institutional-flow strategies need pre-dot-com holdings to separate regime effects from persistent behavior. They pull the information-table rows (CUSIP, issuer, share count, market value) and stack them quarter-by-quarter, joining to post-1999 13F-HR data to build 25+ year manager panels for turnover, concentration, and crowding measures.

Academic finance researchers

Researchers studying 1990s institutional behavior — herding, window dressing, style drift, the rise of indexation — use the dataset to cover the 1994–1999 segment of the manager universe. They rely on manager identity and address fields for classification, filing dates for event-time studies, and holdings tables for portfolio-level statistics.

Financial historians of the asset-management industry

Analysts reconstructing the evolution of bank trust departments, insurance-company subsidiaries, early quant shops, and growth-oriented advisers trace how each category rotated large-cap equity exposure across the decade. They focus on cover-page metadata and the composition of holdings tables over time.

Litigation support and forensic reconstruction teams

Expert witnesses and forensic analysts working 1990s fact patterns — securities class actions, estate and trust disputes, fraud investigations — reconstruct who held a given security at a given quarter-end. The information table, period-of-report date, and 13F-E/A amendments drive exhibit preparation, damages modeling, and expert reports that must cite contemporaneous filings.

Data engineers bridging legacy TXT to modern 13F-HR XML

Engineering teams parse the fixed-column legacy TXT documents and normalize them into the schema used by modern 13F-HR XML. Their work centers on handling older text layouts, aligning CUSIP fields, reconciling 13F-E/A amendments against originals, and mapping manager identifiers to current CIK records. The output is a continuous holdings table with no format break at April 1999.

Index and benchmark research teams

Teams calibrating long-run institutional ownership benchmarks — such as percentage of float held by 13F filers — use the dataset to anchor mid-1990s ownership metrics. They aggregate share counts per CUSIP across all filings in a quarter to validate third-party historical ownership series against primary filings.

Governance and activism historians

Researchers studying the formative years of institutional shareholder activism measure ownership concentration ahead of contested votes and governance events. Manager identity, per-issuer share counts, and quarter-end timing relative to known events support event studies and narrative histories.

LLM and retrieval-system developers on financial archives

Teams building entity-resolution pipelines and retrieval systems over EDGAR use this bounded legacy slice to test parsers against pre-standardization TXT layouts, train manager-name and CUSIP-normalization models, and validate long-horizon retrieval. The closed, small corpus makes it a tractable testbed.

Specific Use Cases

The workflows below reflect how practitioners actually consume the Form 13F-E archive — the only EDGAR-native source for pre-1999 institutional holdings.

Extending factor backtests across the April 1999 boundary

Quantitative researchers parse the fixed-column information table in document-1.txt — CUSIP, VALUE (x$1000), SHARES/PRN AMT, SH/PRN flag, and INVSTMT DSCRETN — for every filing, key each row to metadata.json.periodOfReport, and stack the resulting quarter-end positions beneath a post-1999 13F-HR panel. The output is a continuous 25+ year manager-by-CUSIP holdings matrix used to backtest ownership-based factors (institutional crowding, turnover, 13F concentration) without a regime break at the form-suffix change.

Reconstructing 1990s ownership for securities litigation exhibits

Forensic and expert-witness teams working 1990s class actions, estate valuations, or fraud matters filter records by the cover-page issuer name and 9-character CUSIP, then pull the per-manager SHARES/PRN AMT and SOLE/SHARED/NONE voting-authority columns for the relevant periodOfReport. Amendments (formType = 13F-E/A) are chained to the original via manager CIK plus periodOfReport and ordered by the cover-page AMENDMENT NN label and filedAt, producing exhibits that cite contemporaneous primary filings with defensible revision history.

Empirical papers on 1990s institutional behavior

Academic researchers studying herding, window dressing, and the rise of indexation consume the entities array in metadata.json (CIK, companyName, stateOfIncorporation, sic when present) for manager classification, then compute portfolio-level statistics — Herfindahl concentration, active share, quarter-over-quarter turnover — from the information table. The filedAt timestamp supports event-time studies around 1990s episodes (the 1994 bond-market rout, 1997 Asian crisis, 1998 LTCM window), feeding directly into regression panels and published tables.

Calibrating historical institutional-ownership benchmarks

Index research teams aggregate SHARES/PRN AMT per CUSIP across all filings grouped by periodOfReport to produce a quarterly "percent of float held by 13F filers" series for 1994 through Q1 1999. This primary-source series is used to validate vendor historical-ownership products and to anchor the left tail of long-run institutional-ownership time series that otherwise start in 1999 or 2013.

Building a legacy-to-modern 13F ETL bridge

Data engineers build a column-offset parser against document-1.txt (honoring <PAGE> pagination, the two-line VALUE/SHARES/PUT-CALL/INVSTMT-DSCRETN/MANAGERS/SOLE/SHARED/NONE header grid, and the REPORT SUMMARY checksum line) and emit records that conform to the modern 13F-HR infoTable XML schema. The cover-page "Other Managers" two-digit indices are resolved into CIKs by matching SEC file numbers against metadata.json.entities, yielding a single normalized holdings table whose schema does not change at the April 1999 boundary.

Testbed for EDGAR parser and entity-resolution models

LLM and retrieval-system developers use the bounded corpus as a regression fixture for pre-standardization TXT layouts: evaluating manager-name normalization across the entities array, CUSIP extraction from fixed-width rows, and voting-authority reconciliation (SOLE + SHARED + NONE vs. SHARES/PRN AMT). Because the universe is closed, parser precision and recall can be measured end-to-end without sampling, and the REPORT SUMMARY line provides a row-count and aggregate-value checksum for automated validation.

Dataset Access

Dataset Index JSON API: https://api.sec-api.io/datasets/form-13fe-files.json

This endpoint returns dataset metadata including the name, description, last update timestamp, earliest sample date, total records, total size, covered form types, container format, and included file types. It also returns the full dataset download URL and the list of individual container files, each with its own size, record count, update timestamp, and download URL. Use this endpoint to monitor which containers were updated in the most recent refresh run and to decide which containers to download on a day-by-day basis. This endpoint does not require an API key.

Example response:

          1
          {
        

          2
            "datasetId": "1f13365b-9ae0-699a-8854-855c84c480e6",
        

          3
            "datasetDownloadUrl": "https://api.sec-api.io/datasets/form-13fe-files.zip",
        

          4
            "name": "Form 13F-E Files Dataset",
        

          5
            "updatedAt": "2026-03-21T02:51:19.000Z",
        

          6
            "earliestSampleDate": "1994-01-01",
        

          7
            "totalRecords": 1266,
        

          8
            "totalSize": 28985982,
        

          9
            "formTypes": ["13F-E", "13F-E/A"],
        

          10
            "containerFormat": "ZIP",
        

          11
            "fileTypes": ["TXT", "JSON"],
        

          12
            "containers": [
        

          13
              {
        

          14
                "downloadUrl": "https://api.sec-api.io/datasets/form-13fe-files/1999/1999-04.zip",
        

          15
                "key": "1999/1999-04.zip",
        

          16
                "size": 1843221,
        

          17
                "records": 87,
        

          18
                "updatedAt": "2026-03-21T02:51:19.000Z"
        

          19
              }
        

          20
            ]
        

          21
          }
        

Download Entire Dataset: https://api.sec-api.io/datasets/form-13fe-files.zip?token=YOUR_API_KEY

Downloads the complete Form 13F-E Files dataset as a single ZIP archive containing every container file. This endpoint requires an API key.

Download Single Container: https://api.sec-api.io/datasets/form-13fe-files/1999/1999-04.zip?token=YOUR_API_KEY

Downloads one individual monthly container instead of the full dataset, which is useful for incremental updates or targeted retrieval of a specific period. This endpoint requires an API key.

Frequently Asked Questions

What form does this dataset cover?

The dataset covers Form 13F-E and its amendment counterpart Form 13F-E/A — the EDGAR submission type used by institutional investment managers to file quarterly Section 13(f) holdings reports electronically before April 1999. The /A suffix denotes an amendment that restates, corrects, or supplements a previously filed 13F-E.

What does one record in this dataset represent?

One record corresponds to a single EDGAR submission identified by one accession number — either an original 13F-E quarterly holdings report or a 13F-E/A amendment. On disk, a record is a folder whose name is the 18-digit zero-padded accession number (dashes stripped) and that contains exactly one metadata.json header plus one or more document-<sequence>.txt files carrying the cover page, signature block, other-managers list, and the fixed-column information table.

Who is required to file Form 13F-E?

Institutional investment managers exercising investment discretion over accounts holding at least $100 million in Section 13(f) securities on the last trading day of any month in a calendar year. This population includes registered investment advisers, banks and bank holding companies with trust departments, insurance companies, broker-dealers with discretionary customer accounts, pension plans, endowments, foundations, corporations managing their own portfolios, and other fiduciaries with investment discretion.

How often are records filed, and what time period does the dataset cover?

Form 13F is a quarterly periodic report due within 45 calendar days after the end of each calendar quarter, with holdings reported as of the last day of that quarter. The dataset begins in 1994 at the start of EDGAR voluntary filing and ends in April 1999 when the SEC retired the 13F-E submission type and replaced it with Form 13F-HR (Holdings Report) and Form 13F-NT (Notice).

What file format is the dataset distributed in?

Filings are delivered as ZIP containers holding two file types: a metadata.json header synthesized from the EDGAR SGML submission wrapper, and one or more document-<sequence>.txt bodies. All document bodies are fixed-column ASCII plaintext — Form 13F-E never used HTML or XBRL, and the information table must be parsed with a column-offset/fixed-width strategy rather than an XML or HTML parser.

How does this dataset differ from the modern 13F-HR dataset?

Both cover the same filer population, the same $100 million threshold, the same quarterly cadence, and the same 45-day deadline, but they are separated by era and format. 13F-E is the only EDGAR-native source for 1994 through April 1999 and is delivered in unstructured fixed-column TXT; 13F-HR is the only source from Q2 1999 onward and has been required in structured XML with tagged infoTable elements since 2013. Pre-1999 13F-E also has no HR/NT split, so holdings-versus-notice status must be inferred from the filing text rather than from the form code.

How are 13F-E/A amendments linked back to the original 13F-E?

Each 13F-E/A is a separate EDGAR record with its own accession number and keeps the original period of report rather than creating a new reporting period. Reconstructing an amendment chain requires matching on manager CIK and periodOfReport across records and using the cover-page AMENDMENT NN label together with filedAt to order the amendments, since 13F-E/A encodes amendment intent in free-form cover text rather than in tagged fields.