GEPPML-R: General Equilibrium Analysis with PPML in R

Replication materials for: Anderson, J. E., Larch, M., & Yotov, Y. V. (2018). "GEPPML: General equilibrium analysis with PPML" The World Economy, 41(10), 2750-2782.

📊 Quick Start • 📖 Documentation • 🔬 Methodology • ✅ Validation • 📧 Contact

📋 Overview

This repository provides a complete R implementation of the General Equilibrium Poisson Pseudo-Maximum Likelihood (GEPPML) methodology developed by Anderson, Larch, and Yotov (2018).

Why this package?

🎯 Makes GEPPML accessible to R users who may not have Stata licenses
⚡ Uses modern, high-performance R packages (data.table, fixest)
✅ Produces results identical to the original Stata implementation
📚 Fully documented with comprehensive examples
🔓 Open source (MIT License) for maximum reproducibility

Reference:

Anderson, J. E., Larch, M., & Yotov, Y. V. (2018). GEPPML: General equilibrium analysis with PPML. The World Economy, 41(10), 2750-2782. https://doi.org/10.1111/twec.12664

👨‍🎓 Author Information

Jamiu Olamilekan Badmus

🏛️ Research Intern, Economic and Political Interactions Cluster
United Nations University Institute on Comparative Regional Integration Studies (UNU-CRIS)
🎓 Erasmus Mundus Master in Economics of Globalization and European Integration
📧 Email: jamiubadmus001@gmail.com
🌐 Website: jamiu-olamilekan-badmus
💻 GitHub: @jamiubadmusng

🔍 What is GEPPML?

GEPPML is a methodology for conducting general equilibrium (GE) analysis in international trade using the Poisson Pseudo-Maximum Likelihood (PPML) estimator. It extends the structural gravity model to analyze counterfactual trade policy scenarios by accounting for:

🎯 Baseline

Current trade patterns and trade costs

📊 Conditional GE

Direct policy effects
(holding economic size constant)

🔄 Full GE

Complete equilibrium
(with endogenous adjustments)

Applications:

🤝 Regional Trade Agreements (RTAs)
📉 Trade liberalization policies
🇪🇺 Brexit and other trade shocks
🛤️ Infrastructure improvements affecting trade costs

⭐ Key Features

Feature	Description
✅ Exact Replication	Validated against original Stata code (see VALIDATION.md)
🚀 Modern R Packages	Uses `data.table` (fast data manipulation) and `fixest` (high-performance PPML)
📝 Full Documentation	Comprehensive comments, logging, and user guides
📊 Publication-Ready	Generates country-level indexes and welfare effects for academic research
🔓 Open Source	MIT licensed for maximum accessibility and reproducibility
🔬 Stata Comparison	Includes original Stata code in `GEPPML_stata/` folder for easy comparison

📦 📦 Installation

Prerequisites

This package requires R version 4.0+ and the following R packages:

# Install required packages
install.packages(c("data.table", "fixest", "haven", "dplyr"))

Package	Purpose
`data.table`	Fast data manipulation and aggregation
`fixest`	High-performance PPML estimation with fixed effects
`haven`	Reading Stata .dta files
`dplyr`	Additional data manipulation tools

Clone Repository

git clone https://github.com/jamiubadmusng/GEPPML-R.git
cd GEPPML-R

🚀 🚀 Usage

⚡ Quick Start

Set up your working directory:
```
setwd("path/to/GEPPML_R")
```
Run the main script:
```
source("geppml.R")
```

The script will:

Load and prepare the data
Estimate the baseline gravity model
Compute conditional GE effects
Iterate to convergence for full GE effects
Save all results to the output/ folder
Generate a detailed log file

Input Data

The package uses the original dataset from Anderson et al. (2018): Data File: data/ge_ppml_data_orig.dta
Format: Stata .dta file
Structure: Bilateral trade flows with gravity variables

Variable	Description
`exporter`	Exporting country code (ISO3)
`importer`	Importing country code (ISO3)
`trade`	Bilateral trade flows (US dollars)
`LN_DIST`	Log of bilateral distance (km)
`CNTG`	Contiguity dummy (1 if countries share border)
`BRDR`	Border dummy (1 for international trade, 0 for domestic)

📄 Output

The script generates several output files in the output/ folder:

File	Description
`all_indexes_geppml.csv`	⭐ Main results: Country-level GE indexes (CSV format)
`all_indexes_geppml.rds`	Country-level GE indexes (R binary format)
`full_static_all.rds`	Complete dataset with all GE variables
`imrs_all.rds`	Inward Multilateral Resistance (IMR) indexes
`geppml.log`	Detailed execution log

📊 Key Output Variables

Welfare Effects:

Variable	Description
`rGDP_cndl_ch`	% change in real GDP (conditional GE)
`rGDP_full_ch`	% change in real GDP (full GE)

Trade Effects:

Variable	Description
`tot_exp_cndl_ch`	% change in total exports (conditional)
`tot_exp_full_ch`	% change in total exports (full)

Multilateral Resistance Terms:

Variable	Description
`omr_bsln/cndl/full`	Outward MR (3 scenarios)
`imr_bsln/cndl/full`	Inward MR (3 scenarios)

Other Variables:

Variable	Description
`p_full`	Factory-gate price changes
`acr_bsln`	Domestic absorption share

🔬 Methodology

Three Scenarios

1. Baseline Scenario

Estimates the current trade patterns using a structural gravity equation:

$$X_{ij} = \exp[\beta_1 \ln DIST_{ij} + \beta_2 CNTG_{ij} + \beta_3 BRDR_{ij} + \pi_i + \chi_j] \times \epsilon_{ij}$$

where $\pi_i$ and $\chi_j$ are exporter and importer fixed effects.

2. Conditional GE Scenario

Computes trade effects holding output and expenditure constant, but allowing multilateral resistance terms to adjust:

$$\tilde{X}_{ij}^{CND} = \exp[\pi_i^{CND} + \chi_j^{CND}] \times t_{ij}^{CFL} \times \epsilon_{ij}$$

where $t_{ij}^{CFL}$ represents counterfactual trade costs.

3. Full Endowment GE Scenario

Iteratively solves for the complete general equilibrium by allowing:

Factory-gate prices to adjust
Output and expenditure to change endogenously
All multilateral resistance terms to reach equilibrium

The algorithm iterates until convergence (typically $< 0.001$ for both standard deviation and maximum absolute change in prices).

Key Equations

Outward Multilateral Resistance (OMR): $$\Omega_i = \frac{Y_i \times E_{DEU}}{\pi_i}$$

Inward Multilateral Resistance (IMR): $$\Phi_j = \frac{E_j}{\chi_j \times E_{DEU}}$$

Real GDP: $$\text{Real GDP}_i = \frac{Y_i}{\Phi_i^{1/(1-\sigma)}}$$

where $\sigma$ is the trade elasticity (set to 7 following the literature).

🔄 Differences from Stata Implementation

Feature	Stata (`ppml`)	R (`fixest`)
Fixed effects	Manual creation required	Built-in functionality
Estimation	`ppml` command	`feglm` with Poisson family
Speed	Moderate	Very fast (C++ backend)
Syntax	Stata-specific	Standard R
Flexibility	Limited	Highly flexible

The R implementation produces identical results to the original Stata code while being more accessible and easier to modify.

⚙️ Customization

Changing the Counterfactual

To analyze different trade policy scenarios, modify the trade cost specification:

Example 1: Full Trade Liberalization

# Remove all international borders
dt[, t_ij_ctrf := exp(
  DIST_est * LN_DIST + 
  CNTG_est * CNTG + 
  BRDR_est * 0
)]

Example 2: Regional Trade Agreement

# Create RTA effect
dt[, t_ij_ctrf := exp(
  DIST_est * LN_DIST + 
  CNTG_est * CNTG + 
  BRDR_est * BRDR - 
  0.5 * RTA
)]

Example 3: Reduce Distance Friction

# Reduce distance effect by 50% (e.g., infrastructure improvement)
dt[, t_ij_ctrf := exp(DIST_est * LN_DIST * 0.5 + CNTG_est * CNTG + BRDR_est * BRDR)]

Changing the Reference Country

The default reference country is Germany (DEU). To use a different country:

# Change "DEU" to your desired country code (e.g., "USA" for United States)
dt[exporter == "USA", exporter := "ZZZ"]
dt[importer == "USA", importer := "ZZZ"]

Adjusting the Trade Elasticity

The default trade elasticity is $\sigma = 7$. To change this:

sigma <- 5  # Or any other value based on your research

✅ Replication & Validation

This R implementation has been rigorously validated against the original Stata code. See VALIDATION.md for detailed comparison.

🛠️ How to Replicate

1. Run Stata Version

cd "GEPPML_stata"
do geppml.do

Outputs: all_indexes_geppml.dta

2. Run R Version

source("geppml.R")

Outputs: output/all_indexes_geppml.csv

🎯 Validation Results

Metric	Result
Gravity Coefficients	✅ Identical (6+ decimal places)
Welfare Effects	✅ Max difference: 0.014 percentage points
Convergence	✅ Same iterations (~18-20)
All Countries	✅ Consistent results across 41 countries
Relative Error	✅ < 0.2% (within floating-point precision)

Conclusion: Results are practically identical and suitable for academic publication. Small differences (< 0.02 pp) are within numerical precision tolerance.

📚 Citation

If you use this R implementation in your research, please cite both:

📖 Original Methodology

@article{anderson2018geppml,
  title={GEPPML: General equilibrium 
         analysis with PPML},
  author={Anderson, James E and 
          Larch, Mario and 
          Yotov, Yoto V},
  journal={The World Economy},
  volume={41},
  number={10},
  pages={2750--2782},
  year={2018},
  doi={10.1111/twec.12664}
}

💻 R Implementation

@software{badmus2026geppml,
  author={Badmus, Jamiu Olamilekan},
  title={GEPPML-R: General Equilibrium 
         Analysis with PPML in R},
  year={2026},
  url={https://github.com/jamiubadmusng/
       GEPPML-R}
}

🔗 Related Resources

📊 Stata Comparison: See GEPPML_stata/ folder in this repository

🔧 Troubleshooting

📦 Package 'fixest' not found

install.packages("fixest")

🔄 Convergence not achieved

# Increase maximum iterations (default is 100)
max_iterations <- 200

💾 Memory errors with large datasets

# Increase memory limit (Windows)
memory.limit(size = 16000)  # 16GB

📊 Results differ slightly from Stata

This is expected due to numerical precision differences
Differences should be very small (< 0.02 percentage points)
Ensure you're using the same data and parameter values
See VALIDATION.md for detailed comparison

🤝 Contributing

Contributions are welcome! Please feel free to:

🐛 Report bugs or issues
💡 Suggest enhancements
🔧 Submit pull requests
📚 Share your applications of the code

See CONTRIBUTING.md for guidelines.

📜 License

This project is licensed under the MIT License - see the LICENSE file for details.

🙏 Acknowledgments

James E. Anderson, Mario Larch, and Yoto V. Yotov for developing and sharing the GEPPML methodology
Laurent Bergé for the excellent fixest package

📧 Contact

Questions? Suggestions? Collaboration?

🌟 If you find this package useful, please consider giving it a star on GitHub! 🌟

Last Updated: January 19, 2026

Name		Name	Last commit message	Last commit date
Latest commit History 9 Commits
GEPPML_stata		GEPPML_stata
data		data
output		output
scripts		scripts
.gitignore		.gitignore
CHANGELOG.md		CHANGELOG.md
CITATION.cff		CITATION.cff
CONTRIBUTING.md		CONTRIBUTING.md
FAQ.md		FAQ.md
LICENSE		LICENSE
PACKAGE_INFO.txt		PACKAGE_INFO.txt
PROJECT_STRUCTURE.md		PROJECT_STRUCTURE.md
QUICKSTART.md		QUICKSTART.md
README.md		README.md
REPOSITORY_STRUCTURE.md		REPOSITORY_STRUCTURE.md
VALIDATION.md		VALIDATION.md
geppml.R		geppml.R

License

jamiubadmusng/GEPPML-R

Folders and files

Latest commit

History

Repository files navigation