Appendix Y — Data Card — SEC EDGAR DEF 14A (Proxy Statements)

Provider & what it is

A DEF 14A is the definitive proxy statement a U.S. public company files with the SEC ahead of its shareholder meeting (the “DEF” means definitive, as opposed to a preliminary PRE 14A). It is the single richest public document for corporate governance and executive compensation. Inside one DEF 14A you find: the board’s director nominees and their independence, committee structure (audit, compensation, nominating), the Summary Compensation Table (CEO and named-executive-officer salary, bonus, stock and option awards, total pay), the pay-versus-performance disclosure, beneficial-ownership tables (who owns >5%), the matters up for a vote (say-on-pay, director elections, auditor ratification), and related-party transactions. If a research question involves who runs the firm and how they are paid, this is the source. It is the natural companion to the 10-K, which often incorporates by reference its governance and compensation items from the proxy rather than printing them.

Coverage

EDGAR holds proxy statements from the mid-1990s to the present, filed annually by essentially every U.S.-listed operating company that holds a shareholder meeting. Disclosure content has grown over time by rule: the modern executive-compensation tables date from the SEC’s 2006 compensation-disclosure overhaul, say-on-pay votes arrived with Dodd–Frank (votes from 2011), and the pay-versus-performance table is newer still (filings beginning fiscal 2022 / 2023 proxy season). A panel spanning these dates is therefore not uniformly structured. [CHECK] the exact first proxy season for the pay-versus-performance table before relying on it in early years.

Key identifiers (CIK / accession)

CIK — the filer’s permanent EDGAR company identifier, zero-padded to ten digits in URLs.
Accession number (e.g. 0000320193-24-000007) — the exact handle for one proxy filing; a firm files roughly one DEF 14A per year, but pin it anyway because preliminary (PRE 14A) and additional soliciting materials (DEFA14A) clutter the index.
The filing date and the meeting date inside the body matter for any event study around the annual meeting or a contested vote.
There is no person-level executive identifier in the filing itself; named executives are matched by name (messy), or you cross to a licensed database like Execucomp (WRDS, GMU-only) — a Chapter 7.4 crosswalk.

Access path

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

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

co     = Company("KO")
proxy  = co.get_filings(form="DEF 14A").latest()  # note the SPACE in the form name
text   = proxy.markdown()                          # full proxy text, all tables and narrative

Low-level requests route, polite-sleeping through a loop:

import os, requests, time
HEADERS = {"User-Agent": os.environ["SEC_USER_AGENT"]}
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 == 'DEF 14A'
time.sleep(0.15)                                  # ~7 req/s, under the 10/s ceiling

The reveal: the compensation numbers you want live in HTML tables embedded in the proxy, and — unlike the 10-K’s financial statements — they are not reliably XBRL-tagged, so you usually parse the HTML table directly rather than getting clean structured facts. Pay-versus-performance and a few items have structured tagging in recent years; the older Summary Compensation Table generally does not.

License & note

FREE / public domain. No copyright, no license, no GMU-infrastructure rule — cache, commit, and share freely (size permitting). The compensation tables are public, so a governance/comp study built purely on DEF 14A data is fully re-runnable by any reviewer — the kind of clean reproducibility the data-acquisition chapter prizes. The only caveat is that cleaned, panelized comp data (Execucomp) is a licensed WRDS product; if you use it for convenience, the GMU-infrastructure rule applies to that file, not to the underlying free proxies.

Gotchas

Form-name spaces and siblings. The form is literally DEF 14A (with a space). Do not confuse it with PRE 14A (preliminary), DEFA14A (additional soliciting material), DEFR14A (revised), or DEFC14A (contested/proxy-fight). Filtering loosely pulls in noise.
Tables, not text, carry the data. The Summary Compensation Table is an HTML table with merged cells, footnotes, and multi-year rows; naive text extraction scrambles columns. Parse the table structure, and read the footnotes — option-award “fair value” assumptions and one-time grants live there.
Boilerplate. Governance prose (“the Board believes…”, director-independence language) is heavily templated across firms and dominates a naive bag-of-words.
Incorporation by reference goes both ways. The 10-K may pull comp/governance from the proxy; conversely the proxy may reference the 10-K. Track which document actually contains the item you cite.
Restated / additional materials can supersede figures; pin the accession number and prefer the definitive DEF 14A over later DEFA14A soliciting addenda.

Structure sketch (illustrative — not real values)

A flattened Summary Compensation extract — the shape a comp study builds:

cik	accession_no	fiscal_year	executive	role	salary	stock_awards	option_awards	total_comp
0000021344	0000021344-24-000007	2023	J. Doe	CEO	1,600,000	14,200,000	3,500,000	22,800,000
0000021344	0000021344-24-000007	2023	A. Smith	CFO	850,000	5,100,000	1,200,000	8,400,000
0000021344	0000021344-24-000007	2022	J. Doe	CEO	1,550,000	13,000,000	3,100,000	20,900,000
0000021344	0000021344-24-000007	2022	A. Smith	CFO	820,000	4,700,000	1,000,000	7,600,000

(Illustrative. Names, accession number, and dollar figures are fabricated for shape; verify against the live proxy.)

Where this is used

Chapter 7.2 §7.2.3 (SEC EDGAR) — proxy statements as a governance/compensation source, same fair-access and accession-pinning discipline as 10-Ks.
Governance- and pay-focused research questions (board independence, CEO pay-for-performance, say-on-pay outcomes) and any capstone or extension that links executive compensation or board structure to firm outcomes.
Companion to the 10-K / 10-Q card (Items often incorporated by reference) and to the XBRL card (proxy tables are largely not XBRL, in deliberate contrast).

# Data Card — SEC EDGAR DEF 14A (Proxy Statements) ## Provider & what it is A **DEF 14A** is the *definitive proxy statement* a U.S. public company files with the SEC ahead of its shareholder meeting (the "DEF" means definitive, as opposed to a preliminary **PRE 14A**). It is the single richest public document for **corporate governance and executive compensation**. Inside one DEF 14A you find: the board's director nominees and their independence, committee structure (audit, compensation, nominating), the **Summary Compensation Table** (CEO and named-executive-officer salary, bonus, stock and option awards, total pay), the **pay-versus-performance** disclosure, beneficial-ownership tables (who owns >5%), the matters up for a vote (say-on-pay, director elections, auditor ratification), and related-party transactions. If a research question involves *who runs the firm and how they are paid*, this is the source. It is the natural companion to the 10-K, which often *incorporates by reference* its governance and compensation items from the proxy rather than printing them. ## Coverage EDGAR holds proxy statements from the mid-1990s to the present, filed annually by essentially every U.S.-listed operating company that holds a shareholder meeting. Disclosure content has *grown* over time by rule: the modern executive-compensation tables date from the SEC's **2006** compensation-disclosure overhaul, **say-on-pay** votes arrived with Dodd–Frank (votes from **2011**), and the **pay-versus-performance** table is newer still (filings beginning **fiscal 2022 / 2023 proxy season**). A panel spanning these dates is therefore *not* uniformly structured. **[CHECK]** the exact first proxy season for the pay-versus-performance table before relying on it in early years. ## Key identifiers (CIK / accession) - **CIK** — the filer's permanent EDGAR company identifier, zero-padded to ten digits in URLs. - **Accession number** (e.g. `0000320193-24-000007`) — the exact handle for one proxy filing; a firm files roughly one DEF 14A per year, but pin it anyway because preliminary (PRE 14A) and additional soliciting materials (DEFA14A) clutter the index. - The **filing date** and the meeting date inside the body matter for any event study around the annual meeting or a contested vote. - There is no person-level executive identifier in the filing itself; named executives are matched by name (messy), or you cross to a licensed database like Execucomp (WRDS, GMU-only) — a Chapter 7.4 crosswalk. ## Access path **Free and public, no key.** Same **fair-access rule** as all of EDGAR: every request carries a `User-Agent` header with your name and email, and you stay under **10 requests/second** aggregated across scripts. Missing User-Agent returns `403`. ```python from edgar import Company, set_identity set_identity("Maya Rodriguez maya@gmu.edu") # fair-access identity; read from env in real code co = Company("KO") proxy = co.get_filings(form="DEF 14A").latest() # note the SPACE in the form name text = proxy.markdown() # full proxy text, all tables and narrative ``` Low-level `requests` route, polite-sleeping through a loop: ```python import os, requests, time HEADERS = {"User-Agent": os.environ["SEC_USER_AGENT"]} 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 == 'DEF 14A' time.sleep(0.15) # ~7 req/s, under the 10/s ceiling ``` The reveal: the compensation *numbers* you want live in HTML tables embedded in the proxy, and — unlike the 10-K's financial statements — they are **not reliably XBRL-tagged**, so you usually parse the HTML table directly rather than getting clean structured facts. Pay-versus-performance and a few items have structured tagging in recent years; the older Summary Compensation Table generally does not. ## License & note **FREE / public domain.** No copyright, no license, no GMU-infrastructure rule — cache, commit, and share freely (size permitting). The compensation tables are public, so a governance/comp study built purely on DEF 14A data is fully re-runnable by any reviewer — the kind of clean reproducibility the data-acquisition chapter prizes. The only caveat is that *cleaned, panelized* comp data (Execucomp) is a licensed WRDS product; if you use it for convenience, the GMU-infrastructure rule applies to *that* file, not to the underlying free proxies. ## Gotchas - **Form-name spaces and siblings.** The form is literally `DEF 14A` (with a space). Do not confuse it with **PRE 14A** (preliminary), **DEFA14A** (additional soliciting material), **DEFR14A** (revised), or **DEFC14A** (contested/proxy-fight). Filtering loosely pulls in noise. - **Tables, not text, carry the data.** The Summary Compensation Table is an HTML table with merged cells, footnotes, and multi-year rows; naive text extraction scrambles columns. Parse the table structure, and read the footnotes — option-award "fair value" assumptions and one-time grants live there. - **Boilerplate.** Governance prose ("the Board believes…", director-independence language) is heavily templated across firms and dominates a naive bag-of-words. - **Incorporation by reference goes both ways.** The 10-K may pull comp/governance from the proxy; conversely the proxy may reference the 10-K. Track which document actually *contains* the item you cite. - **Restated / additional materials** can supersede figures; pin the accession number and prefer the definitive DEF 14A over later DEFA14A soliciting addenda. ## Structure sketch (illustrative — not real values) A flattened Summary Compensation extract — the shape a comp study builds: | cik | accession_no | fiscal_year | executive | role | salary | stock_awards | option_awards | total_comp | |-----|--------------|-------------|-----------|------|--------|--------------|---------------|------------| | 0000021344 | 0000021344-24-000007 | 2023 | J. Doe | CEO | 1,600,000 | 14,200,000 | 3,500,000 | 22,800,000 | | 0000021344 | 0000021344-24-000007 | 2023 | A. Smith | CFO | 850,000 | 5,100,000 | 1,200,000 | 8,400,000 | | 0000021344 | 0000021344-24-000007 | 2022 | J. Doe | CEO | 1,550,000 | 13,000,000 | 3,100,000 | 20,900,000 | | 0000021344 | 0000021344-24-000007 | 2022 | A. Smith | CFO | 820,000 | 4,700,000 | 1,000,000 | 7,600,000 | (*Illustrative.* Names, accession number, and dollar figures are fabricated for shape; verify against the live proxy.) ## Where this is used - **Chapter 7.2** §7.2.3 (SEC EDGAR) — proxy statements as a governance/compensation source, same fair-access and accession-pinning discipline as 10-Ks. - Governance- and pay-focused research questions (board independence, CEO pay-for-performance, say-on-pay outcomes) and any capstone or extension that links executive compensation or board structure to firm outcomes. - Companion to the **10-K / 10-Q** card (Items often incorporated by reference) and to the **XBRL** card (proxy tables are largely *not* XBRL, in deliberate contrast).