PyMarkup-estimator 0.2.0

A Python toolkit for estimating firm-level markups using production function-based marginal cost recovery.

PyMarkup-estimator

A Python toolkit for estimating firm-level markups using production function-based marginal cost recovery.

• PyPI package: https://pypi.org/project/PyMarkup-estimator/
• GitHub repository: https://github.com/immortalsRDJ/PyMarkup-estimator
• Free software: MIT License
• Documentation: https://PyMarkup-estimator.readthedocs.io

Key Features

• Three Production Function Estimators: Compare results across multiple methods

  • Wooldridge IV: IV/GMM with lagged COGS as instrument, rolling 5-year windows by industry (main method, addresses simultaneity bias)
  • Cost Share: Direct accounting approach assuming perfect competition (quick benchmark)
  • ACF: Ackerberg-Caves-Frazer two-stage GMM with productivity proxy (robustness check)

• Flexible Configuration: YAML config files or programmatic Python API

• Industry-Level Estimation: Rolling windows with configurable minimum observation thresholds

• Clean Modular Architecture: Separation between core logic, estimators, pipeline orchestration, and I/O

• Rich Output Options:

  • Save to CSV, Parquet, or Stata formats
  • Plot aggregate markup time series
  • Compare methods with summary statistics
  • Full type hints and validation with Pydantic

• Command-Line Interface: Run complete pipeline with a single command

• Intermediate Results: Optionally save panel data, elasticities, and markups at each step

Quick Start

Installation

pip install PyMarkup-estimator

Note: The package name is PyMarkup-estimator, but you import it as PyMarkup:

from PyMarkup import MarkupPipeline  # Import name stays the same!

Python API

from PyMarkup import MarkupPipeline, PipelineConfig, EstimatorConfig

# Configure pipeline
config = PipelineConfig(
    compustat_path="data/compustat.dta",
    macro_vars_path="data/macro_vars.xlsx",
    estimator=EstimatorConfig(
        method="wooldridge_iv",  # or "cost_share", "acf", "all"
        iv_specification="spec2",
        window_years=5
    ),
    output_dir="output/"
)

# Run pipeline
pipeline = MarkupPipeline(config)
results = pipeline.run()

# Save and analyze results
results.save("output/", format="csv")
comparison = results.compare_methods()
fig = results.plot_aggregate()

Command Line

# Using config file
pymarkup estimate --config config.yaml

# Direct parameters
pymarkup estimate --method wooldridge_iv \
    --compustat data/compustat.dta \
    --macro-vars data/macro_vars.xlsx \
    --output results/

How It Works

The MarkupPipeline orchestrates a 3-step workflow:

1. Data Preparation: Load Compustat panel, merge macro variables, deflate to real terms, apply percentile trimming
2. Elasticity Estimation: Estimate production function output elasticities (θ) by industry-year using selected method(s)
3. Markup Calculation: Compute firm-level markups as markup = θ_cogs / cost_share

Requirements

• Python 3.10+
• pandas, numpy, scipy
• linearmodels (for IV/2SLS estimation)
• statsmodels
• pydantic (for configuration validation)
• typer (for CLI)

Development

# Install with development dependencies
pip install -e ".[test]"

# Run tests
pytest

# Run quality checks
ruff check .
ruff format .

TODO

• Decomposition