talsan / ceopay

source and structure executive compensation data from SEC Edgar, with optional aws integration (s3 + athena)

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Source & Structure Executive Compensation from SEC Edgar Filings

Process Overview

Generalized for any Edgar Document (10-K, 10-Q, DEF 14A, etc.)

  1. Download Master Index Files: process critical lookup index for each Edgar filing (company name, filing date, filing url, etc.)
  2. Extract Filing Headers: additional filing metadata (contents within filing's <SEC_HEADER> tag)
  3. Download Filings: in this case, the annual proxy filings (formtype="DEF 14A")

Specific to Executive Compensation (Form DEF 14A)

  1. Detect Compensation Table: Random Forest classifier finds the compensation table out of 100s of random/noisy tables in the raw DEF14A document
  2. Parse Table Contents: Messy and inconsistent raw HTML tables get translated into structured JSON objects with Salary details (total, base, bonus, options, etc.)

Features & Options

  • Decoupled into functional units:
    1. a self-contained event processor (e.g. idx_downloader.py)
    2. a batch process that queues-and-invokes a series of events (e.g. batch_idx_downloader.py).
  • Multiple output options: Local directories or S3
  • Multiprocessing can be toggled on/off via config.py
  • Designed to work as a script (via CLI) or as an imported module within Python

Process Architecture

Process Architecture

Process Details

1. Download Master Index Files

idx_downloader.py
Inputs:
usage: idx_downloader.py [-h] [--local_output LOCAL_OUTPUT] yyyyqq

downloads individual quarterly master index files from SEC edgar

positional arguments:
  yyyyqq                <yyyyqq> formatted year-quarter pair (eg 202001) to download from edgar

optional arguments:
  -h, --help            show this help message and exit
  --local_output LOCAL_OUTPUT
                        where to send output on local machine; defaults to 's3', which uploads to the config.Aws.OUPUT_BUCKET defined in
                        config.py)
Outputs:

Output Example
S3 naming convention: <config.Aws.OUPUT_BUCKET>/masteridx/year=2020/qtr=1.txt
Local naming convention: ./ceopay/data/masteridx/year=2020/qtr=1.txt

2. Extract Filing Headers

hdr_extractor.py
Inputs:
usage: hdr_extractor.py [-h] [--local_output LOCAL_OUTPUT] yyyyqq formtype

extract additional metadata (contents within each filing's <SEC-HEADER> tags) across all filings within an individual master index file
(filtered on a given 'formtype')

positional arguments:
  yyyyqq                <yyyyqq> formatted year-quarter pair (eg 202001) to download from edgar
  formtype              eg "DEF 14A", "10-K", "10-Q", etc

optional arguments:
  -h, --help            show this help message and exit
  --local_output LOCAL_OUTPUT
                        where to send output on local machine; defaults to 's3', which uploads to the config.Aws.OUPUT_BUCKET defined in
                        config.py)
Outputs:

Output Example
S3 naming convention: <config.Aws.OUPUT_BUCKET>/filing_metadata/formtype=def14a/year=2020/qtr=1.txt
Local naming convention: ./ceopay/data/filing_metadata/formtype=def14a/year=2020/qtr=1.txt

About

source and structure executive compensation data from SEC Edgar, with optional aws integration (s3 + athena)


Languages

Language:Python 100.0%