Form U5B Files Dataset

The Form U5B Files Dataset is a closed historical archive of Form U5B and Form U5B/A registration statements filed with the SEC under Section 5 of the Public Utility Holding Company Act of 1935 (PUHCA 1935). Each record in the dataset is one EDGAR submission — either an initial registration statement (U5B) or an amendment (U5B/A) — identified by its 18-digit SEC accession number and packaged with the original filer-submitted documents plus a structured metadata.json. The filer is always the U.S.-domiciled holding-company parent that became subject to mandatory registration because it owned, controlled, or held with power to vote 10% or more of the voting securities of an electric utility company or a retail gas utility company. The dataset spans EDGAR full-text coverage of the form, beginning January 1, 1995 and ending February 2006, when the Energy Policy Act of 2005 repealed PUHCA 1935 and the form was discontinued. Records are distributed as monthly ZIP archives containing TXT, JSON, HTML, and PDF files.

Dataset APIs

Dataset Files

What This Dataset Contains

The dataset captures the full SEC EDGAR record of registered-holding-company registration under PUHCA 1935. Form U5B is the substantive registration statement that Section 5 of PUHCA required of any corporation that became a registered holding company; Form U5B/A is the amendment variant used to supplement, correct, or replace specific Items of a previously filed U5B. Together they form a closed, non-growing corpus: the form was retired in February 2006 after the Energy Policy Act of 2005 repealed PUHCA, which fixes a hard upper bound, and the lower bound is January 1995, the start of EDGAR full-text coverage for this form.

The unit of observation is the accession, not the registrant. A holding company that amends its registration statement multiple times appears as multiple distinct records, one per accession, each anchored to its own filedAt timestamp and forming an independent point in the amendment chain. For each accession the dataset preserves the structured filing-level metadata that EDGAR exposes on the filing-detail page together with every textual document the filer submitted — primary registration statement plus exhibits — passed through verbatim under their original EDGAR filenames. Image attachments from the original submission (.jpg and .gif files used for organizational charts and corporate logos) are intentionally omitted; all other textual material is retained. Because PUHCA addressed integrated public-utility systems, the form's content is descriptive and structural rather than financial in the income-statement sense; financial information typically arrives by incorporation by reference to the registrant's already-filed Form 10-K and other periodic reports rather than as restated figures.

The dataset is distributed as a hierarchy of ZIP archives, one archive per filing month at the path <year>/<year>-<month>.zip, with file extensions TXT, JSON, HTML, and PDF.

Content Structure of a Single Record

What one record represents

A single record in the Form U5B Files Dataset is one EDGAR submission of either Form U5B or Form U5B/A, identified by its 18-digit SEC accession number. The record packages the complete filing as it was accepted by EDGAR — the filing-level metadata together with every textual document the filer submitted — with image attachments excluded. The unit of observation is the accession, not the registrant: a holding company that amends its registration statement multiple times appears as multiple distinct records, one per accession, each anchored to its own filedAt timestamp and forming an independent point in the amendment chain.

Container and on-disk layout of one record

The dataset is distributed as a hierarchy of ZIP archives, one archive per filing month at the path <year>/<year>-<month>.zip. Inside each ZIP, files live under a top-level <year>-<month>/ directory, and each filing sits in its own subfolder named with the 18-character accession number with dashes stripped (for example, accession 0001047469-04-014246 becomes folder 000104746904014246, matching EDGAR's URL convention). The on-disk record is therefore one folder, deterministically locatable from the dashed accession number plus the filedAt year and month, which makes the dataset trivially indexable without scanning the archives.

Inside an accession folder there are exactly two kinds of files. First, a single metadata.json holding the structured filing-level summary. Second, one or more original EDGAR-submitted documents passed through verbatim under their original filenames — names are not normalized and reflect whatever the filer's printer produced (e.g. a2135003zu5ba.htm, where the numeric component is the EDGAR Filing Workstation job number and the suffix encodes the form variant). Image attachments from the original EDGAR submission, such as .jpg and .gif files used for organizational charts and corporate logos, are intentionally omitted; all other textual material is retained.

The file extensions found in the dataset are TXT, JSON, HTML, and PDF. In practice this resolves to: .htm/.html for the primary U5B or U5B/A statement and any HTML-formatted exhibits; .txt for plain-text exhibits such as legal opinions, fixed-width subsidiary tables, and incorporated correspondence; occasional .pdf files when the original submission included rasterized or typeset exhibits; and the always-present metadata.json.

The metadata.json component

Every accession folder carries a metadata.json that mirrors what EDGAR exposes on the filing-detail page, recast as structured JSON. Its salient fields are:

• formType — U5B or U5B/A, the only two values present in the dataset.
• accessionNo — the canonical dashed accession number.
• filedAt — ISO 8601 timestamp with offset, recording the EDGAR acceptance moment.
• description — human-readable form title, including amendment language for U5B/A (e.g. Form U5B/A - Registration statement filed under [Section 5]: [Amend]).
• linkToFilingDetails, linkToTxt, linkToHtml — URLs pointing back to the primary document, the full SGML submission text, and the EDGAR filing index page on sec.gov.
• linkToXbrl — uniformly empty; Form U5B was discontinued before any XBRL tagging mandate could apply to it.
• documentFormatFiles[] — an enumeration of the submitted documents. Each entry carries sequence, size, documentUrl, description, and type. The first sequence is the primary U5B or U5B/A document; subsequent entries are exhibits in submission order; the trailing entry, whose sequence and type are a single-space sentinel, denotes the EDGAR full-submission .txt bundle.
• dataFiles[] — an empty array for this form type; the field is reserved for ancillary data attachments that U5B does not carry.
• entities[] — one entry per entity associated with the filing. Fields include act (the statutory act, typically 35 for PUHCA), cik, fileNo, irsNo, companyName annotated with the role in parentheses (e.g. (Filer)), type, sic (industry classification string, frequently 4900 Electric, Gas & Sanitary Services for the utility holding companies in scope), filmNo, and tickers[].
• id — an opaque 32-character hexadecimal record identifier.

Because Form U5B is a free-text registration statement, no XBRL or financial-fact extraction is layered on top of the document; the metadata captures filing-level identification and points back at the documents, while the substantive disclosure remains inside the HTML/TXT/PDF documents themselves.

The document component and its SGML wrapper

The primary U5B or U5B/A document is preserved exactly as EDGAR stored it — that is, as a fragment inside the SGML <DOCUMENT> envelope that EDGAR uses inside the full-submission text file. The on-disk file therefore opens with header tags and closes with their counterparts:

          1
          <DOCUMENT>
        

          2
          <TYPE>U5B/A
        

          3
          <SEQUENCE>1
        

          4
          <FILENAME>a2135003zu5ba.htm
        

          5
          <DESCRIPTION>U5B/A
        

          6
          <TEXT>
        

          7
          <HTML> ... </HTML>
        

          8
          </TEXT>
        

          9
          </DOCUMENT>
        

The <TYPE>, <SEQUENCE>, <FILENAME>, and <DESCRIPTION> lines reproduce the EDGAR document-header metadata; the body of the registration statement is the complete <HTML> document nested inside <TEXT>. Each separately-submitted exhibit is wrapped in its own <DOCUMENT> block with its own <TYPE> (commonly EX-99 for exhibits in the U5B G-1/G-2/G-3 series, plus other letter-coded types depending on the exhibit class) and is stored on disk as its own file. Plain-text exhibits use the same envelope but with raw text rather than HTML in the <TEXT> block. PDF exhibits are uuencoded inside the EDGAR submission TXT but are written to disk as the decoded .pdf payload.

Internal structure of the U5B / U5B/A document

The HTML body inside the SGML envelope follows the form-instructions ordering and typically contains the following components in sequence:

• A cover page bearing the full form title (FORM U5B / REGISTRATION STATEMENT FILED PURSUANT TO SECTION 5 OF THE PUBLIC UTILITY HOLDING COMPANY ACT OF 1935, or its AMENDMENT NO. n TO FORM U5B analogue for U5B/A), the exact legal name of the registrant, the principal executive office address, and notice-recipient information.
• The numbered Form U5B Items as prescribed by the form instructions. These cover, in order: the corporate identity of the registrant; identification of the chief accounting officer; the holding-company system's organizational structure; tabular disclosure of each direct and indirect subsidiary, including its position in the system, the nature and book value of its assets, and the geographic and interstate scope of its operations; descriptions of franchises held, securities outstanding, and material intercompany contracts; and pointers to financial information that is incorporated by reference from the registrant's Form 10-K or other periodic reports rather than restated inline. In amendments (U5B/A), Items appear selectively — only the Items being modified are present, often introduced by language explaining what is being replaced, added, or removed relative to the prior filing.
• Signature blocks, attestation, and notarization tables identifying the corporate officers signing on behalf of the registrant, frequently rendered as HTML tables.
• A verification block in which the signing officer formally attests, under oath, to the truth of the foregoing.
• An index of exhibits, traditionally lettered (Exhibit A for the charter and bylaws, Exhibit B for indentures and intercompany contracts, Exhibit C for opinions of counsel, Exhibit D for tax allocation agreements, Exhibit E for service contracts, Exhibit F for organizational charts, Exhibit G for financial statements and schedules, with numeric suffixes such as G-1, G-2, G-3 distinguishing successive items within a class). Each listed exhibit corresponds to a separate <DOCUMENT> block in the EDGAR submission and to a separate file on disk.

Interleaved through the HTML body are filing-printer comment markers of the form <!-- ZEQ.=1,SEQ=2,EFW="2135003",CP="...",DN="1",CHK=...,FOLIO='1',FILE='DISK...',USER='...',CD='28-APR-2004;12:22' -->. These are RR Donnelley / EDGAR composition-system artifacts encoding printer identity, composition timestamps, and source-file paths. They have forensic value but should be stripped before any content extraction or text analysis.

What is included and what is excluded

Each record includes, for one accession: the structured metadata.json; the primary U5B or U5B/A registration statement preserving its EDGAR SGML wrapper; and every accompanying exhibit document the filer submitted in textual form (HTML, plain text, or PDF). Excluded are image attachments from the original EDGAR submission — graphics, scanned organizational-chart bitmaps, corporate seals, and similar .jpg/.gif files — and any XBRL artifacts, which never existed for this form. Content that the filing only references by incorporation (Form 10-K excerpts, prior periodic filings, court orders) is also outside the record; only the U5B body's pointers to that material are preserved, not the referenced material itself.

Changes in required content over the dataset's lifespan

Form U5B's content requirements were defined by PUHCA Section 5 and the SEC's Form U5B instructions, both of which were stable across the dataset's January 1995 to February 2006 window. The scope of disclosure — corporate identity, chief accounting officer, subsidiary tabulation, asset and operational description, exhibit index, verification — remained constant. The dataset therefore does not exhibit the episodic Item-additions or renumbering seen in other long-lived forms. The single material disclosure-regime change is structural rather than internal: PUHCA was repealed by the Energy Policy Act of 2005, the form was discontinued in February 2006, and there are no records after that point.

Changes in data format over the dataset's lifespan

Across the 1995–2006 window the EDGAR presentation of Form U5B did evolve. Early-1990s filings were predominantly plain ASCII text wrapped in the SGML <DOCUMENT> envelope, with tabular subsidiary disclosures rendered as fixed-width text and exhibits delivered as .txt files; this is the source of the TXT extension in the dataset's file-type set. From the late 1990s onward, filers and their printers migrated to HTML for the primary statement and for many exhibits, producing the <HTML>-inside-<TEXT> shape that dominates later years and accounts for the prominence of .htm/.html files. PDF exhibits appear sporadically when filers submitted scanned or typeset documents alongside the structured text. The SGML <DOCUMENT> wrapper itself is invariant across the entire span — every record, regardless of era, presents its documents wrapped in the same envelope, and only the payload inside <TEXT> shifts from raw ASCII to HTML over time.

Interpretation and extraction notes

Several nuances matter when working with these records.

• Amendment chains. U5B/A amendments are independent records keyed by their own accession; reconstructing the current state of a registrant's registration statement requires walking the chain of U5B and U5B/A filings for that CIK in chronological order, since amendments commonly replace specific Items rather than restate the whole filing.
• Incorporation by reference. Much of the substantive financial information is present only by reference to the registrant's Form 10-K and other periodic filings; the U5B body itself frequently contains pointers rather than restated figures, and any analysis that needs balance-sheet detail must follow those references out into adjacent EDGAR datasets.
• SGML wrapper around HTML. Document files retain the SGML <DOCUMENT> wrapper and must be treated as SGML-wrapped HTML or text rather than as clean standalone documents. An HTML parser must be fed only the <TEXT> body, with the surrounding <TYPE>/<SEQUENCE>/<FILENAME>/<DESCRIPTION> lines either stripped or read separately as document-level metadata.
• Printer comment markers. RR Donnelley / EDGAR composition comments (<!-- ZEQ=...,SEQ=...,EFW=... -->) embedded in HTML bodies should be removed before text extraction; they are not part of the disclosure.
• HTML-entity encoding in metadata. entities[].sic values in metadata.json carry HTML-entity-encoded ampersands (e.g. Electric, Gas & Sanitary Services) and require decoding for human display.
• Sentinel entry in documentFormatFiles[]. The array always ends with an entry whose sequence and type are a single space — this entry represents the EDGAR full-submission TXT bundle. Downstream code must not assume sequence is always numeric or type is always non-empty.
• Deterministic on-disk paths. Any record's location is reconstructible from the dashed accession number and the filedAt year and month, which makes targeted retrieval feasible without scanning the monthly archives.
• Mixed-format exhibits in one accession. A single record can mix HTML, plain-text, and PDF exhibits; consumers must dispatch on file extension rather than assume a uniform body format across an accession.

Who Files or Publishes This Dataset, and When

Who files

The filer is the holding-company parent that is becoming, or has become, a registered holding company under Public Utility Holding Company Act of 1935. Specifically, a Form U5B filer is a corporation that owned, controlled, or held with power to vote 10% or more of the voting securities of an electric utility company or a retail gas utility company, and that was therefore subject to mandatory registration under Section 5 of PUHCA.

The filer is always the parent registrant, not its subsidiaries. Utility operating companies and nonutility affiliates are described inside the U5B in tabular form (organizational structure, assets, interstate operations) but they do not file U5B themselves.

Out of scope for this filer population:

• Holding companies claiming a Section 3 exemption (intrastate or otherwise). These filed Form U-3A-2 or related exemption forms and never appear here.
• Utility operating subsidiaries.
• Foreign parents above the registered U.S. holding company. The U5B is filed by the U.S.-domiciled registrant inside the chain of control, not by an upstream foreign parent.

The Form U5B population is therefore narrower than "utility holding companies": it is only those that actually entered the registered regime.

When the record is created

Form U5B is event-driven, not periodic. A filing is triggered by the registration event itself:

• A company that became subject to PUHCA registration first filed Form U5A (the short notification of registration).
• It then had 90 days from obtaining registered status to file Form U5B as the substantive registration statement.

The 90-day clock is anchored to the registration event, not to a fiscal year end. There is no recurring U5B obligation; once filed, ongoing PUHCA disclosure was carried by other U-forms (notably Form U5S, the annual report for registered holding companies).

Form U5B/A amendments are filed by the same registrant whenever the facts in the original registration statement change or require correction (corporate restructurings, change of principal executive office, change of chief accounting officer, addition or disposition of subsidiaries, factual corrections). Amendments have no fixed cadence and do not reset the 90-day clock. A registered system typically appears in the dataset as one initial U5B plus zero or more U5B/A filings.

The dataset itself is a closed historical window. EDGAR coverage begins January 1, 1995. The form was discontinued in February 2006 after the Energy Policy Act of 2005 repealed PUHCA 1935 (effective February 8, 2006) and shifted residual oversight to FERC under PUHCA 2005. No new U5B records exist after that date.

Important distinctions

• U5A vs. U5B: U5A is the short notification that triggers registered status; U5B is the substantive registration statement due within 90 days after. Companion filings, but only U5B and U5B/A are in this dataset.
• U5B vs. U5S: U5B is the one-time registration statement. U5S is the recurring annual report for registered holding companies. Do not conflate them.
• U5B vs. exemption forms: Registered holding companies file U5B; exempt holding companies file U-3A-2 and similar. The two populations do not overlap.
• U5B vs. Exchange Act reporting: Many registrants were also Exchange Act issuers filing 10-K, 10-Q, 8-K, and Section 16 forms. U5B is a PUHCA-only registration document and does not substitute for, or duplicate, Exchange Act periodic reporting.
• Filer vs. entities described: The legal filer is the registering parent. Subsidiaries listed inside the U5B (utility and nonutility) are subjects of disclosure, not filers.

How This Dataset Differs From Similar Datasets or Filings

Form U5B sits inside a narrow, closed corner of SEC disclosure: filings under PUHCA 1935. The most useful comparison targets are the other PUHCA U-series forms, the Exchange Act periodic reports filed by the same parents, and the post-2006 FERC regime that replaced this framework. S-series registration statements and beneficial ownership forms are not relevant: U5B is structural-registration disclosure for a utility holding company system, not a securities offering or ownership filing.

Form U5A — Notification of Registration

The trigger document for the same regime. A holding company filed U5A to notify the SEC that it had become a holding company under PUHCA 1935; U5B followed within 90 days as the substantive registration statement.

Difference: U5A is procedural and minimal — identification of the new holding company and the date status was acquired. U5B is substantive — tabular disclosure of subsidiaries, asset descriptions, organizational charts, officers, and interstate operations. U5A captures the event; U5B captures the resulting structural snapshot.

Form U5S — Annual Report by Registered Holding Companies

The annual periodic counterpart to U5B. After registration, registered holding companies filed U5S each year to update operations, intercompany transactions, service-company arrangements, and subsidiary structure.

Difference: U5B is a one-time registration baseline; U5S is recurring and richer in financial detail (intercompany loans, service-company billings, transaction flows). For longitudinal analysis use U5S; for the registration-moment snapshot use U5B. They are complements.

Form U-1 — Application or Declaration

The transactional approval filing under PUHCA 1935, used to obtain SEC authorization for specific corporate actions (issuances, acquisitions, dispositions, intra-system financings) under Sections 6, 7, 9, 10, or 12.

Difference: Form U-1 is event-driven, transaction-specific, frequent, and legally argumentative. U5B is structural, one-time, descriptive, and tabular. U-1 documents what registered systems did; U5B documents what those systems were.

Form 10-K filed by the same parent

Many U5B filers (e.g., Southern Company, AEP, Entergy, Allegheny Energy) also filed 10-Ks because their securities were publicly traded.

Difference: 10-K is Exchange Act, investor-focused, narrative-heavy, and organized around Items (Business, Risk Factors, MD&A, audited financials, controls). U5B is PUHCA-specific, regulator-focused, and tabular about utility-system architecture, interstate operations, and the legal status of each subsidiary. 10-K does not document PUHCA concepts such as integrated public-utility system status, single-area requirements, or system service company arrangements; U5B contains no risk factors, MD&A, audited consolidated statements, or compensation disclosure. 10-K coverage also continues after 2006; U5B does not.

FERC PUHCA 2005 filings (post-February 2006)

The Energy Policy Act of 2005 repealed PUHCA 1935 effective February 8, 2006 and replaced it with PUHCA 2005, administered by FERC. Holding companies now file FERC Form 65 (notification), 65-A (waivers), and 65-B (exemptions) under 18 C.F.R. Part 366.

Difference: jurisdictional and substantive. FERC filings live in FERC eLibrary, not EDGAR, and are outside the SEC dataset universe. The FERC regime is materially lighter — it eliminated most substantive approvals (the U-1 workload) and replaced detailed structural registration with simple notification. U5B cannot be extended forward via FERC, and FERC filings cannot be backfilled with U5B content.

Boundary summary

U5B is distinct on five axes the comparisons isolate: registration-stage (vs. U5A notification), one-time (vs. U5S annual), structural (vs. U-1 transactional), regulator-focused on utility-system architecture (vs. 10-K investor-focused financials), and SEC/PUHCA 1935 (vs. FERC/PUHCA 2005). Within the PUHCA 1935 family, U5B is the canonical structural-snapshot document: pair it with U5S for ongoing operations, with U-1 for transaction activity, and move to Exchange Act or FERC sources for any post-2006 view of the same parents.

Who Uses This Dataset

The Form U5B Files Dataset is consulted by a narrow set of professionals reconstructing the structure of U.S. utility holding companies as they existed at the moment of PUHCA repeal.

Energy regulatory counsel

Public utility lawyers use the registrant identification, subsidiary tables, and U5B/A amendment trail to establish what was on file when PUHCA 1935 was repealed. This baseline supports books-and-records disputes, affiliate-transaction reviews, and grandfathered-status arguments under PUHCA 2005 administered by FERC.

FERC compliance and rate analysts

Compliance staff at investor-owned utility holding companies cross-reference the historical subsidiary perimeter against current corporate trees when scoping FERC Form 60, centralized service company filings, and EQR submissions. The tabular subsidiary disclosure (organizational placement, assets, interstate footprint) drives affiliate-transaction and cost-allocation analyses.

Utility M&A diligence teams

Bankers and transaction counsel advising on utility, transmission, and generation deals use the organizational charts and asset summaries to trace target genealogy back to a registered holding company system. The records surface legacy entities carrying residual liabilities, pre-deregulation rate-base assets, or interstate service contracts previously approved by the SEC.

Forensic accountants and litigation support

In ratepayer, securities, and environmental cost-recovery matters with PUHCA-era roots, forensic accountants treat U5B filings as a contemporaneous, attested view of holding-company assets and operations at a fixed historical date. Originals and U5B/A amendments are read together to identify what registrants later corrected.

Energy economists and IO researchers

Economists studying utility consolidation, vertical integration, and the 1935-to-2005 regulatory transition use subsidiary tables to build longitudinal panels of system size, geographic reach, and asset mix. Pre-repeal disclosures are compared against post-repeal corporate structures to measure divestiture and ring-fencing activity.

Entity-resolution data scientists

Teams building energy-sector reference data use the registrant name, principal office address, chief accounting officer, and subsidiary list as authoritative anchors for legacy holding-company graphs. The records reconcile cleanly against later 10-K Exhibit 21 lists, FERC filings, and state PUC records, supporting CIK-to-entity crosswalks.

Regulatory archivists

Information-management staff at law libraries and policy research institutions index the corpus as a finite, non-growing record of a discontinued form. The metadata files and original EDGAR submissions support reference requests and citations in regulatory histories.

Policy researchers

Energy policy analysts cite specific U5B disclosures as primary evidence in retrospectives on PUHCA repeal, cross-subsidization, and consumer-protection design under PUHCA 2005. Outputs include white papers, regulatory comment letters, and testimony.

Credit analysts on legacy utility debt

Analysts covering long-tenor utility issuers consult U5B filings when indenture covenants, ring-fencing provisions, or rating methodologies reference former registered-holding-company status. The subsidiary list confirms which operating utilities sat inside the registered system at a given date.

RAG and retrieval-system developers

Engineering teams building energy-regulatory retrieval systems ingest the bounded corpus (TXT, HTML, PDF with JSON metadata) as a curated source for questions on registered holding companies, subsidiary structures, and the PUHCA transition. Structured registrant metadata supports reliable citation to accession numbers.

Specific Use Cases

The use cases below reflect concrete workflows against the fixed body of Form U5B and U5B/A registration statements filed under PUHCA 1935 between January 1995 and February 2006.

• Reconstructing the registered-system perimeter at PUHCA repeal. Pull the last U5B or U5B/A on file for each registrant before February 8, 2006 and parse the subsidiary tables into a structured list of operating utilities, intermediate holding entities, and nonutility affiliates. This produces the baseline corporate tree used in PUHCA 2005 grandfathered-status arguments, FERC Form 60 scoping, and centralized service-company filings.

• Walking U5B/A amendment chains for forensic and litigation work. For a given CIK, order the U5B and U5B/A accessions chronologically by filedAt and diff the Items present in each amendment against the prior filing. The output identifies which subsidiaries, asset descriptions, or officer attestations the registrant later corrected, supporting ratepayer disputes, environmental cost-recovery matters, and securities litigation tied to pre-repeal disclosures.

• Building a legacy holding-company entity graph. Extract registrant name, CIK, principal executive office, chief accounting officer, IRS number, and SIC from metadata.json, then join to the parsed subsidiary tables inside the HTML or TXT body. The resulting parent-subsidiary edges anchor entity-resolution crosswalks against later 10-K Exhibit 21 lists, FERC eLibrary records, and state PUC dockets.

• Tracing M&A target genealogy to a registered system. For a utility, transmission, or generation asset under diligence, search the subsidiary tables across the corpus to find the U5B that originally listed the entity, then read the surrounding asset description and interstate-operations disclosure. This identifies legacy intercompany contracts, pre-deregulation rate-base assets, and residual affiliate obligations carried into the modern target.

• Constructing a longitudinal panel of utility holding-company structure. Across the corpus, tabulate subsidiary count, asset-class mix, and geographic and interstate footprint per registrant-year. The panel supports econometric work on utility consolidation, vertical integration, and divestiture activity straddling the 1935-to-2005 regulatory transition.

• Powering a bounded RAG corpus on PUHCA 1935. Strip the SGML <DOCUMENT> wrapper and RR Donnelley printer comments from each primary document, chunk the Items and exhibits, and index alongside metadata.json fields. The closed, non-growing corpus delivers reliable accession-level citations for questions on registered holding companies, integrated public-utility system status, and exhibit-class contents (charters, indentures, opinions, tax allocation agreements, service contracts, organizational charts).

Dataset Access

The Form U5B Files Dataset is accessible through three methods: a JSON metadata endpoint, a full dataset archive download, and individual container file downloads.

Dataset Index JSON API: https://api.sec-api.io/datasets/form-u5b-files.json

This endpoint returns dataset-level metadata (name, description, last updated timestamp, earliest sample date, total record count, total size, form types covered, container format, and file types) along with the full dataset download URL and a list of all container files. Each container entry includes its size, record count, last updated timestamp, and direct download URL. This endpoint does not require an API key. It can be polled to monitor which containers have been refreshed in recent runs and to decide which containers to fetch incrementally.

Example response:

          1
          {
        

          2
            "datasetId": "1f13365b-9ae0-6a45-85d7-034105aae211",
        

          3
            "datasetDownloadUrl": "https://api.sec-api.io/datasets/form-u5b-files.zip",
        

          4
            "name": "Form U5B Files Dataset",
        

          5
            "updatedAt": "2026-04-16T08:41:14.910Z",
        

          6
            "earliestSampleDate": "1995-01-01",
        

          7
            "totalRecords": 370,
        

          8
            "totalSize": 4707485,
        

          9
            "formTypes": ["U5B", "U5B/A"],
        

          10
            "containerFormat": "ZIP",
        

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

          12
            "containers": [
        

          13
              {
        

          14
                "downloadUrl": "https://api.sec-api.io/datasets/form-u5b-files/2005/2005-12.zip",
        

          15
                "key": "2005/2005-12.zip",
        

          16
                "size": 184320,
        

          17
                "records": 6,
        

          18
                "updatedAt": "2026-04-16T08:41:14.910Z"
        

          19
              }
        

          20
            ]
        

          21
          }
        

Download Entire Dataset: https://api.sec-api.io/datasets/form-u5b-files.zip?token=YOUR_API_KEY

Downloads the complete archive containing all Form U5B and U5B/A filings from January 1995 through the form's discontinuation in February 2006. This endpoint requires an API key.

Download Single Container: https://api.sec-api.io/datasets/form-u5b-files/2005/2005-12.zip?token=YOUR_API_KEY

Downloads a single monthly container archive instead of the full dataset, useful for targeted retrieval of filings from a specific month. This endpoint requires an API key.

Frequently Asked Questions

What forms does this dataset cover?

The dataset contains exactly two form types: U5B, the substantive registration statement filed under Section 5 of PUHCA 1935, and U5B/A, the amendment variant used to supplement, correct, or replace specific Items of a previously filed U5B. No other form types appear in the dataset.

What does one record in this dataset represent?

One record is a single EDGAR submission of Form U5B or Form U5B/A, identified by its 18-digit SEC accession number. The record contains a structured metadata.json plus every textual document the filer submitted (primary statement and exhibits), with original filenames preserved. Image attachments such as .jpg and .gif files from the EDGAR submission are intentionally excluded.

Who was required to file Form U5B?

The filer is the U.S.-domiciled holding-company parent that owned, controlled, or held with power to vote 10% or more of the voting securities of an electric utility company or a retail gas utility company, and that was therefore subject to mandatory registration under Section 5 of PUHCA 1935. The filer is always the parent registrant; subsidiaries listed inside the U5B are subjects of disclosure, not filers. Holding companies claiming a Section 3 exemption are not in this population — they filed Form U-3A-2 instead.

When was a U5B filing due, and how often were amendments filed?

Form U5B was event-driven, not periodic: a registrant had 90 days from obtaining registered status to file U5B, after first filing the short Form U5A notification. There was no recurring U5B obligation. Form U5B/A amendments were filed whenever facts changed or required correction (corporate restructurings, change of principal executive office or chief accounting officer, addition or disposition of subsidiaries) and have no fixed cadence.

What time period does the dataset cover?

The dataset is a closed historical window. EDGAR coverage begins January 1, 1995 (the start of EDGAR full-text coverage for the form) and ends in February 2006, when the Energy Policy Act of 2005 repealed PUHCA 1935 (effective February 8, 2006) and the form was discontinued. No new U5B records exist after that date.

How is the dataset distributed and what file formats are inside?

The dataset is distributed as a hierarchy of monthly ZIP archives at the path <year>/<year>-<month>.zip. Inside each archive, every filing sits in its own subfolder named with the dashes-stripped accession number. File extensions found in the dataset are TXT, JSON, HTML, and PDF: .htm/.html for the primary statement and HTML exhibits, .txt for plain-text exhibits, occasional .pdf for typeset or scanned exhibits, and an always-present metadata.json.

How does Form U5B differ from Form U5S and Form U-1?

U5B is the one-time structural registration statement filed at the moment a holding company entered the PUHCA 1935 registered regime. Form U5S is the recurring annual report filed by registered holding companies after registration, and is richer in financial detail (intercompany loans, service-company billings, transaction flows). Form U-1 is the transactional approval filing used to obtain SEC authorization for specific corporate actions such as issuances, acquisitions, or intra-system financings. U5B documents what registered systems were; U5S documents how they operated annually; U-1 documents what they did transactionally.