Skip to content

Director compensation data

This dataset converts ISS director compensation records into a public firm-year panel and attaches Compustat and Revelio firm identifiers.

Legacy workflow

This builder still requires the retired mapping_revelio_compustat.parquet input. It is retained for reference, not as a runnable supported workflow.

flowchart TB
    iss[(ISS director compensation)]
    comp[(Processed annual Compustat)]
    mapping[(Revelio-Compustat mapping)]
    build[Clean pay and link firms]
    panel[(Director compensation data)]

    iss --> build
    comp --> build
    mapping --> build
    build --> panel

    classDef source fill:#e8eef7,stroke:#52739e
    classDef process fill:#fff0cc,stroke:#a66b00
    classDef data fill:#e2f2e7,stroke:#3b7a4b
    class iss,comp,mapping source
    class build process
    class panel data
Inputs
Dataset Location
ISS director compensation Vendor ISS delivery under /data/full/raw/
Processed annual Compustat /data/full/processed/compustat_annual.parquet
Revelio--Compustat mapping /data/full/processed/mapping_revelio_compustat.parquet
Build
Component Location
Builder shared/data/build_datasets/director_comp_panel.py
Runtime hct-tower-model-cuda-… (CPU execution)
Entry point python -m shared.data.build_datasets.director_comp_panel
Output
Dataset File / location Grain Key and variables
Director compensation panel iss_director_annual_comp.parquet in /data/full/raw/iss/ Director-firm-year Key: cik, participantid, year
Variables include gvkey, rcid, cash_comp, equity_comp, other_comp, total_comp, log_total_comp

The builder removes fiscal-year stubs, constructs cash, equity, other, total, and log compensation, joins Compustat by CIK and year, and attaches a primary Revelio rcid through gvkey. Coverage by year is written beside the main file.

python -m shared.data.build_datasets.director_comp_panel
python -m shared.data.build_datasets.director_comp_panel --inspect

There is currently no project Slurm wrapper for this builder. Run the direct command in a CPU job using the model CUDA image.

The current data support firm-level links only because the ISS extract lacks director names. The main panel is source-derived and does not depend on HCT scores, so it is documented as reusable processed data despite its legacy /data/full/raw/iss/ location.