Compustat¶
Purpose¶
Compustat is used as a licensed comparison dataset, particularly for its annual advertising-expense field (xad). It is not used to construct the SEC XBRL disclosure measures. Comparing the two sources helps distinguish the presence of an XBRL-tagged disclosure from the availability of a standardized vendor field.
Raw input dataset¶
The raw input is Compustat North America Fundamentals Annual, commonly accessed as the comp.funda table. The project currently uses annual data only.
The project-level extract is stored as:
/data/compustat_advertising_annual.parquet
Variables currently requested¶
| Group | Variables |
|---|---|
| Firm, industry, and fiscal-year identifiers | cik, conm, tic, sic, naics, gind, gsubind, fyear, fyr |
| Advertising and SG&A | xad, xsga |
| Operating and investment controls | sale, at, capx, ppent, xrd, cogs, oibdp, ib, oancf |
| Capital structure, financing, and payout controls | dlc, dltt, lt, ceq, csho, prcc_f, sstk, dltis, dv |
The returned annual records also carry Compustat's gvkey and reporting-end date (datadate), which are used to identify and align observations. The project adds xad_available and xsga_available to flag nonmissing values; these are derived availability flags rather than fields from the raw Compustat table.
xad is the main comparison measure: a missing value means that an advertising-expense value is unavailable in this Compustat extract, not that advertising expense was zero.
Joining to XBRL data¶
cik is the principal identifier for linking the Compustat extract to the XBRL data. The data sources have different reporting conventions and timing, so joins should also align the reporting end date (datadate versus ddate) and should not assume that a Compustat xad observation maps one-to-one to a single XBRL tag.
License and sharing¶
Compustat is licensed data. Do not commit the extracted Parquet file to the public repository or share it with collaborators who lack the appropriate WRDS/Compustat access. Each authorized user should create their own extract with their own credentials.