Skip to content

Legacy exploratory goodwill and target MQ

This analysis asks whether an acquired target's pre-deal MQ is associated with the goodwill reported by the acquirer.

Superseded for the current M&A result

This exploratory builder uses the legacy curated CSV. The current Table 9 pipeline is documented in Deals and starts from the reprocessed XBRL deal panel.

Manuscript pipeline incomplete

The Python builder creates the analysis dataset and embeds coefficient summaries in JSON, but it does not save fitted models or write a LaTeX table.

flowchart LR
    data[(Goodwill and target MQ)]
    fit[[Run pooled and FE regressions]]
    json[(Coverage and regression summary JSON)]
    latex[/Planned LaTeX table/]

    data --> fit --> json -.-> latex

    classDef input fill:#eee5f5,stroke:#76528c,stroke-width:2px
    classDef code fill:#f3f3f3,stroke:#666,stroke-width:2px
    classDef support fill:#fff0cc,stroke:#a66b00,stroke-width:2px
    classDef tableGap fill:#e2f2e7,stroke:#a66b00,stroke-width:2px,stroke-dasharray:5 5
    class data input
    class fit code
    class json support
    class latex tableGap
Inputs
Dataset Location
Goodwill and target MQ data Built from the goodwill extract and configured MQ source
Build
Component Location
Builder shared/data/build_datasets/goodwill_target_mq.py
Runtime Model CUDA image (hct-tower-model-cuda-…)
Entry point python -m shared.data.build_datasets.goodwill_target_mq
Output
Dataset File / location Grain and variables
Goodwill-target MQ result Deal-level Parquet output and goodwill_target_mq_summary.json under shared/data/build_datasets/outputs/ unless overridden Deal-level coverage and regression summaries; no manuscript table yet

The current builder estimates log goodwill on target MQ measured at t-1 and over the three pre-deal years. Its specification grid includes pooled, year, industry, industry-plus-year, acquirer, and acquirer-plus-year models, with additional acquirer-size variants. Standard errors are heteroskedastic-robust or clustered by acquirer when available.

cd /hct-tower
python -m shared.data.build_datasets.goodwill_target_mq \
  --goodwill-path shared/data/external/goodwill_xbrl_20102022.csv \
  --fuzzy

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

A manuscript-ready implementation should separate estimation from data construction, persist model or tidy numeric results, and write a table fragment under /artifacts/<coverage>/analyses/.