Replication Package for "Europe Falling Behind: Structural Transformation and Labor Productivity Growth Differences Between Europe and the U.S."
Description
This dataset contains the complete replication package for Buiatti, Duarte, and Sáenz (2026), "Europe Falling Behind: Structural Transformation and Labor Productivity Growth Differences Between Europe and the U.S.", Journal of International Economics. The paper develops a multi-sector structural transformation model to study the divergence in labor productivity growth between Europe and the United States since the mid-1990s, and quantifies the roles of sectoral productivity growth and labor reallocation across sectors through counterfactual experiments under three model variants (closed economy, open economy with exogenous trade, and open economy with endogenous trade). The package is self-contained: running `python code/master.py` reproduces every figure and table in the paper and its online appendix end-to-end in approximately 7-12 minutes on modern Apple Silicon (Python 3.11 with pandas 2.x, or Python 3.8 with pandas 1.5). All 19 pipeline steps are orchestrated from master.py, and README.pdf provides a per-figure and per-table replication map. Contents. code/ — all Python scripts (model calibration, European tests, closed-economy and trade counterfactuals, figure and table generators, and utility modules for shift-share decomposition and LP-TFP correlation analysis); data/ — analysis-ready panels (EU KLEMS 2023, Penn World Table 10.0, OECD GDP per hour, OECD ICIO trade panels) and the raw source files under data/raw_data/ (including the ~180 MB EU KLEMS Growth Accounts file for the Table 4 TFP correlation analysis); requirements.txt — Python dependencies; README.md / README.pdf — setup and replication instructions; LICENSE — MIT. Requirements. Python 3.8 or newer with the packages in requirements.txt; a LaTeX installation (MacTeX, TeX Live, or MiKTeX) for matplotlib's text rendering. Source and versioning. This dataset mirrors the v1.0 tagged release of the GitHub repository https://github.com/jbduarte/europe-falling-behind-replication.
Files
Steps to reproduce
1. Prerequisites. Python 3.8 or newer (Python 3.11 recommended for faster runtime) and a LaTeX installation (MacTeX on macOS; TeX Live on Linux; MiKTeX on Windows) for matplotlib's text rendering. 2. Unpack and install dependencies. unzip europe-falling-behind-replication-v1.0.zip cd europe-falling-behind-replication pip install -r requirements.txt 3. Run the full pipeline. cd code python master.py This executes all 19 steps end-to-end in approximately 7 minutes on Python 3.11 with pandas 2.x, or approximately 12 minutes on Python 3.8 with pandas 1.5, on modern Apple Silicon (MacBook Pro or iMac with M2 chip or newer). 4. Outputs. All figures and tables are written to output/figures/ and output/tables/. Step 19 (generate_paper_outputs.py) copies them under paper-consistent names: figure_1.pdf through figure_A5.pdf, table_1.xlsx through table_A10.tex. 5. Per-figure / per-table replication map. README.pdf, Section "Output Mapping," documents which script produces each paper figure and table. For example, Figure 3 and Table 2 are produced by counterfactuals.py (Step 3); Figure 6 and Table 6 by trade_counterfactuals_endogenous.py (Step 9); Table 4 by utils/corr_lp_tfp_klems.py (Step 18). 6. Note on Step 18 (Table 4, LP-TFP correlation). This step requires data/raw_data/growth accounts.csv (~180 MB), included in this Mendeley package. If working from the GitHub mirror at https://github.com/jbduarte/europe-falling-behind-replication, this file must be downloaded separately from the EU KLEMS 2023 Growth Accounts release at https://euklems-intanprod-llee.luiss.it/ and placed at data/raw_data/growth accounts.csv. If the file is missing, Step 18 is skipped gracefully and the remaining 18 steps complete normally. 7. Individual step execution (optional). Each of the 19 steps may also be run standalone (for example, python counterfactuals.py), provided earlier steps in the dependency chain have been executed first. master.py handles the ordering automatically.
Institutions
- University of South CarolinaSouth Carolina, Columbia
- NOVA School of Business and EconomicsLisbon, Carcavelos