Appendix X — Data Card — SEC EDGAR 8-K (Material Events)

Provider & what it is

The 8-K is the SEC’s “something just happened” form. Where the 10-K and 10-Q are scheduled annual and quarterly reports, the 8-K is filed on demand whenever a material corporate event occurs — and it must usually be filed within four business days of the event. That timeliness is exactly what makes it the natural raw material for an event study: the filing date is a sharp, machine-readable timestamp for “new information hit the market.” Each 8-K is organized by numbered Items that name the event type — Item 2.02 (results of operations / earnings release), Item 5.02 (departure or appointment of directors and officers), Item 1.01 (entry into a material agreement), Item 2.01 (completion of an acquisition), Item 8.01 (other events), and so on. This is the source behind Capstone 4 — SEC 8-K Text Classification, where you classify the event type from the text and study the return reaction around the filing.

Coverage

EDGAR holds 8-Ks from the mid-1990s to the present, filed by essentially every U.S.-listed operating company. A typical large firm files anywhere from a handful to dozens of 8-Ks a year. The modern Item-numbering scheme (the 1.0x / 2.0x / 5.0x structure) dates from the SEC’s August 2004 8-K overhaul; filings before that use an older, coarser item taxonomy, so a sample spanning 2004 is not coded consistently. [CHECK] the exact pre-2004 item codes if your study reaches back that far.

Key identifiers (CIK / accession)

  • CIK — the filer’s permanent EDGAR company identifier, zero-padded to ten digits in URLs.
  • Accession number (e.g. 0000320193-23-000045) — the exact handle for one 8-K. Because a firm files many 8-Ks per year, the accession number, not “the 2023 8-K,” is the only unambiguous reference.
  • Item numbers inside the filing are the event identifiers (2.02, 5.02, …); they are the labels your classifier in Capstone 4 either predicts or validates against.
  • The filing date (and, for some events, an event date in the body) is the timestamp your event window centers on.

Access path

Free and public, no key. Same fair-access rule as all of EDGAR: every request carries a User-Agent header naming you and your email, and you stay under 10 requests/second aggregated across scripts. A missing User-Agent returns 403.

from edgar import Company, set_identity
set_identity("Sam Okafor sam@gmu.edu")           # fair-access identity; read from env in real code

co = Company("TSLA")
eightks = co.get_filings(form="8-K")              # all 8-Ks for the company
recent  = eightks.head(20)
text    = recent[0].markdown()                    # full text of one 8-K, items included

Or the low-level requests route, sleeping between calls so a loop stays polite:

import os, requests, time
HEADERS = {"User-Agent": os.environ["SEC_USER_AGENT"]}
for cik in cik_list:
    url = f"https://data.sec.gov/submissions/CIK{int(cik):010d}.json"
    r = requests.get(url, headers=HEADERS, timeout=30); r.raise_for_status()
    # filter r.json()['filings']['recent'] where form == '8-K' ...
    time.sleep(0.15)                              # ~7 req/s, under the 10/s ceiling

For finding event filings by language, the full-text search endpoint (EFTS, 2001–present) returns filings whose text matches a query — useful to assemble a candidate set before you pull and classify. edgartools exposes the same throttling-aware machinery, so prefer it for large jobs.

License & note

FREE / public domain. No copyright, no license, no GMU-infrastructure restriction — cache and commit freely (size permitting). Note for the AI module: when you feed 8-K text to an LLM classifier, the filings are public, but your API calls to a commercial LLM are not free of obligations — log them, disclose AI use per the capstone’s responsible-use checklist (Reading Guide Pack 6), and never paste licensed data (none here) into a vendor. The reveal: the data is free, but reproducibility of a stochastic LLM label is not automatic — pin the model version and temperature, and cache the model’s output alongside the raw filing.

Gotchas

  • Multiple events per filing. One 8-K can carry several Items at once (e.g. 2.02 and 9.01 exhibits). Your label is therefore multi-label, not single-label — a classic Capstone 4 modeling decision, not a bug.
  • Earnings releases hide in Exhibit 99.1. Item 2.02 8-Ks usually point to the actual press release as Exhibit 99.1; the body of the 8-K is a one-line pointer, and the content lives in the exhibit. Pull the exhibit, not just the cover.
  • Boilerplate cover language (“furnished, not filed” disclaimers, forward-looking-statement safe-harbor paragraphs) repeats across firms and will dominate a naive bag-of-words unless you strip it.
  • Amendments (8-K/A) correct or supplement an earlier 8-K; pin the accession number and decide which you study.
  • Filing-time vs. event-time. The four-business-day window means the filing date can lag the event; for a tight event study, read the body for the event date and be explicit about which clock you use (a look-ahead trap).

Structure sketch (illustrative — not real values)

A parsed 8-K event table, the shape Capstone 4 builds:

cik accession_no filing_date items exhibit_991 event_label
0001318605 0001318605-23-000045 2023-10-18 2.02, 9.01 yes earnings
0001318605 0001318605-23-000051 2023-11-02 5.02 no exec_change
0001318605 0001318605-23-000060 2023-12-07 1.01, 9.01 no material_agreement
0001318605 0001318605-24-000003 2024-01-24 8.01 no other
0001318605 0001318605-24-000009 2024-02-12 2.01 yes acquisition_complete

(Illustrative. Accession numbers, dates, and item combinations are fabricated for shape; verify against live filings.)

Where this is used

  • Capstone 4 — SEC 8-K Text Classification — classify event type from 8-K text with out-of-sample validation, then study the return reaction around the filing date.
  • Week 6 AI module (Ch 6.5, the AI Co-Pilot) and Reading Guide Pack 6 — LLM text classification with OOS validation, leakage audits, and the responsible-use/disclosure checklist that Capstone 4 must satisfy.
  • Chapter 7.2 §7.2.3 (SEC EDGAR) — the access pattern, fair-access rule, and accession-number pinning this card details.