SEC Form 4, 2026 Q2: issuer-as-owner filings, flag spellings and filing timing
Aervik Labs (operated by EAC Ventures, LLC)
https://aerviklabs.com/research/sec-form4-2026q2/

LICENSE
  CC BY 4.0 -- https://creativecommons.org/licenses/by/4.0/
  Free to use and redistribute, including commercially, with attribution.

SUGGESTED CITATION
  Aervik Labs. "SEC Form 4, 2026 Q2: issuer-as-owner filings, flag spellings and filing timing." Measurement of the SEC Insider Transactions Data Set 2026q2 and 600 EDGAR Form 4 XML documents, read 2026-09-14. https://aerviklabs.com/research/sec-form4-2026q2/ (CC BY 4.0)

WHAT THIS IS
  Three measurements of SEC Form 4 filings with filing dates 2026-04-01 to 2026-06-30
  (2026 Q2). (a) and (b) were read from SEC on 2026-09-14 (UTC); (c) was read on 2026-09-15 from the same
  data set file (same SHA-256).

  (a) Issuer as reporting owner. Of the 4,497 companies that were the issuer on at least one
      Form 4 in the quarter, 35 also appeared as a reporting owner on a Form 4 for a
      different issuer: 80 filings out of 49,832 Form 4s. For those companies, filtering
      their own EDGAR filing list to form type 4 returns Form 4s whose issuer is another company.
  (b) Flag spellings. How isDirector, isOfficer, isTenPercentOwner, isOther and aff10b5One are
      written (1/0 or true/false) in a seeded random sample of 600 Form 4 XML documents.
  (c) Filing timing. Business days from each transaction date to the filing date, for every Form 4
      in the quarter (49,636 filings and 107,116 transaction rows measured), overall and by
      transaction code, and what the form's own timing fields (the transactionTimeliness value, the
      deemed execution date and the transaction form type) show for the rows filed more than two
      business days after their transaction date.
      An interval here is a measured distance between two dates. It is not a finding that any filing was late, and nothing here says that any filing, filer or issuer broke a rule.

  Subjects: Companies (issuer CIKs) and filings (accession numbers). No natural person is named, listed or counted as a subject.
  For (c): Filings and transaction rows, counted in aggregate. No issuer, reporting owner, CIK, accession number or document filename appears.

FILES
  companies-issuer-as-reporting-owner.csv
    One row per company (35 rows), sorted by CIK.
    company_cik                   the company's SEC Central Index Key
    company_name_as_issuer        the issuer name the SEC data set records for this CIK on its Form 4s
                                  in the quarter, verbatim
    cross_issuer_form4_filings    Form 4s filed in the quarter whose issuer is a different company and
                                  that list this CIK as a reporting owner
    own_issuer_form4_filings      Form 4s filed in the quarter with this CIK as the issuer
    listed_form4_filings          own + cross: the quarter's Form 4s listed under this CIK
    cross_issuer_share_of_listed  cross / listed, four decimals
    Appearing in this file reflects how EDGAR indexes a filing. It says nothing about the company.

  sample-flag-spellings.csv
    One row per sampled document (600 rows), sorted by sample_rank.
    accession_number              the filing's EDGAR accession number
    sample_rank                   position in the draw (see METHOD)
    draw_key_sha256               lowercase hex SHA-256 of "<seed>|<accession_number>"
    issuer_cik                    <issuerCik> as written in the XML
    document_type                 <documentType> as written in the XML
    schema_version                <schemaVersion> as written in the XML
    reporting_owners              number of <reportingOwner> elements
    For each ELEMENT in isDirector, isOfficer, isTenPercentOwner, isOther (up to one per reporting owner)
    and aff10b5One (one per document):
      ELEMENT_values              the element's trimmed text for each occurrence, in document order, joined
                                  with "|"; empty means the element is not in the document
      ELEMENT_word                occurrences written true or false
      ELEMENT_numeric             occurrences written 1 or 0
      ELEMENT_other               occurrences written anything else
    relationship_flags_spelling   numeric, word or numeric+word, over the four relationship flags
    all_five_spelling             the same, over all five elements
    mixes_relationship_flags      1 if both spellings occur among the four relationship flags
    mixes_all_five                1 if both spellings occur across all five elements
    xml_bytes                     size of the XML document text as fetched (UTF-8 bytes)
    xml_sha256                    SHA-256 of the XML document text as fetched
    matches_data_set              1 if the XML's issuer CIK, document type, reporting-owner count and
                                  raw aff10b5One value all equal the SEC data set's
    To open a document: https://www.sec.gov/Archives/edgar/data/<issuer_cik>/<accession_number
    without dashes>/index.json lists the filing's files; the sampled document is its top-level .xml.
    No owner names and no document filenames are included.

  summary.json
    Headline figures for (a) and (b), sources with read times, sample parameters, license and citation.

  lag-summary.json
    Every filing-timing figure on the page (c): population, per-filing distribution, per-row shares,
    the transactionTimeliness values, deemed execution dates, transaction form type 5, rows with no
    visible timing marker, Form 4/A, codes with at least 1,000 rows, thin tails, verification,
    calendar checks, the specification and rule text quoted on the page, and prior work.

  lag-by-code.csv
    One line per table and transaction code (35 lines). table is nonderivative (Table I),
    derivative (Table II) or both. A code with at least 100 measured rows in that table has its own
    line; the rest are pooled into code "other" (pooled_codes lists each code and its row count), and
    code "all" is the whole table. Business days (bd) run from the row's transaction date to the filing
    date. Counts only, no identifiers.
    rows                                 Form 4 transaction rows
    rows_measured                        rows minus those whose transaction date is after the filing date
    rows_transaction_date_after_filing   counted and not measured
    filings                              distinct filings behind the measured rows (blank when pooled)
    median_bd, p90_bd, p99_bd            nearest rank over measured rows
    bd_0_2, bd_3_5, bd_6_30, bd_over_30  measured rows in each business-day range
    rows_over_2_bd                       measured rows more than 2 business days after the transaction
    filings_with_rows_over_2_bd          distinct filings behind those rows (blank when pooled)
    most_rows_over_2_bd_in_one_filing    the largest number of those rows in one filing (blank when pooled)
    Of rows_over_2_bd:
    deemed_date_reported                 rows with a deemed execution date
    within_2_bd_of_deemed_date           rows 2 or fewer business days after the deemed execution date
    transaction_form_type_5              rows with transaction form type 5
    timeliness_blank, timeliness_e,      rows whose transactionTimeliness is empty, E or L
    timeliness_l
    no_visible_marker                    rows with no deemed execution date, no transaction form type 5
                                         and no timeliness value
    thin_tail                            1 when rows_over_2_bd >= 20 and filings_with_rows_over_2_bd < 50:
                                         that code's tail describes a handful of filings (blank on the
                                         other and all lines)

  README.txt
    This file.

METHOD
  Sources: SEC Insider Transactions Data Set 2026q2 (https://www.sec.gov/files/datastandardsinnovation/data/insider-transactions-data-sets/2026q2_form345.zip,
  Last-Modified 2026-07-09T15:16:30Z, 11498860 bytes, SHA-256 11f1b2bbbdcbe6347a34437c02d04202fda0eca1dbb023726e4b56504b802e27);
  EDGAR full index for 2026 QTR2 (https://www.sec.gov/Archives/edgar/full-index/2026/QTR2/master.gz); filing indexes and XML documents
  from EDGAR Archives; company filing lists from data.sec.gov.

  (a) Companies = distinct issuer CIKs over the data set's Form 4 filings. A cross-issuer filing is a
      Form 4 with at least one reporting-owner CIK that is one of those companies and is not the
      filing's own issuer. CIKs are compared as integers. Computed three independent ways -- an
      in-memory join of the SUBMISSION and REPORTINGOWNER tables, a SQL join of the same tables, and
      a join against EDGAR's quarterly full index in place of REPORTINGOWNER -- which returned the same
      companies and the same (filing, company) pairs. EDGAR's full index lists the same Form 4 and
      Form 4/A accession sets as the data set. On 2026-09-14, all 80 filings were present with
      form 4 in the companies' filing lists on data.sec.gov.
  (b) Population: the data set's 49,832 Form 4 accessions (Form 4/A not sampled). Seed
      "aervik-sec-form4-2026q2-20260914". Key = SHA-256 of "<seed>|<accession_number>"; sort
      ascending; the first 600 are the sample. A filing that failed to resolve would have been replaced
      by the next rank; none failed (0 excluded). Each element's text is trimmed; true/false is
      "word", 1/0 is "numeric". Two independent parsers (regular expressions over the raw text; an XML
      tree parser over schema positions) agreed on all 18 compared fields for all 600 documents.
  (c) Source: the same data set zip, read 2026-09-15T02:33Z (Last-Modified 2026-07-09T15:16:30Z,
      SHA-256 11f1b2bbbdcbe6347a34437c02d04202fda0eca1dbb023726e4b56504b802e27).
      Filings whose SEC data set document type is 4. Form 4/A amendments are measured separately and are in no Form 4 figure.
      Business day: not a Saturday, not a Sunday, and not a date in OPM's federal holiday schedule tables. Holidays from https://www.opm.gov/policy-data-oversight/pay-leave/federal-holidays/
      (read 2026-09-15T02:39Z); OPM's 2026 dates equal SEC's EDGAR Calendar (11 dates).
      Interval: the number of business days after the transaction date, up to and including the filing date. Per filing: from the earliest transaction date across both tables of the filing to its filing date.
      Percentiles: nearest rank, the value at rank ceil(p x n).
      Two independent programs (one reading the extracted tables, one reading the zip directly with its
      own holiday parse and business-day count) agreed on all 17,523 compared values. 30 filings
      (20 whose earliest transaction was more than 45 calendar days before filing, 10 within 3 calendar days) were compared
      with their own submission text on EDGAR: 30 of 30 matched on the filing date, every transaction date,
      deemed execution date, transactionTimeliness and transactionFormType.

ROUNDING
  Sample shares are whole percentages: at n = 600 the worst-case margin of error at 95% confidence is
  +/-4.0 percentage points. Full-quarter shares are exact counts shown to one decimal,
  and so are all filing-timing shares.

LIMITATIONS
  - One quarter. Nothing here says other quarters look the same.
  - (a) is a floor: it counts only companies that were a Form 4 issuer in the quarter. Widening the
    universe to any Form 3, 4 or 5 issuer gives 38 of 5,065.
  - A company's recent filing list on data.sec.gov usually spans more than one quarter; the quarter
    figure is not a per-lookup rate.
  - SEC disclaims the accuracy of the data set's extraction. Mitigated by the full-index coverage
    match and the 600-document XML cross-check; not eliminated for fields not checked.
  - (b) is a sample of 600. Zero documents mixing spellings within the relationship flags puts the
    95% upper bound at 0.64%; it does not prove mixing never happens.
  - Form 4/A is excluded from (a) and from the sample. Including 4/A on both sides of (a) gives
    35 of 4,508 companies and 81 filings.
  - (c) is a filing-date cohort: every Form 4 filed in the quarter, whatever its transaction date.
  - (c) An interval over two business days is not by itself a late filing. The rules deem some
    execution dates later than the trade date (17 CFR 240.16a-3(g)(2) and (g)(3)), allow transactions
    reportable on Form 5 to be reported early on Form 4 (17 CFR 240.16a-3(g)(5)), and date a filing
    by EDGAR's cutoff (17 CFR 232.13(a)(4)); the data set carries dates, not times.
  - (c) EDGAR's daily indexes were checked against the calendar for 2025-01-01 to 2026-06-30; no daily
    index exists for 3 weekdays counted as business days, and treating them as closed changes 0 bucket
    counts and no percentile by more than 3 business days. Before 2011-01-01 only weekends are excluded.
  - (c) Footnotes and remarks were not read. Rows with no visible marker are not a count of late filings.
  - (c) Form 4/A is measured separately (672 amendments, median 13 business days) and is in no Form 4 figure.

Not investment advice.