Data and Code for: Useful Work and the Productivity Paradox (OECD, 1980–2020)

This dataset contains the replication data and code for the manuscript “Useful Work and the Productivity Paradox: Heterogeneous Output Elasticities, Structural Change, and Biophysical Regimes in OECD Economies, 1980–2020.”

The repository provides the merged country-year panel, the main Jupyter Notebook used to reproduce the empirical analyses, and additional figure-rendering scripts used to generate the publication-ready figures. The deposited CSV contains the full constructed panel with 18 OECD economies and 591 country-year observations. The main estimation sample excludes Germany and Ireland because of structural breaks unrelated to the research question: German reunification and Irish IP-transfer GDP distortions. After these exclusions, the main ARDL estimation sample covers 16 OECD economies and 519 estimation observations.

The data combine variables from Penn World Table version 11.0 and the Country-Level Primary-Final-Useful (CL-PFU) energy and exergy database v1.2. The core variables are Country, Year, Y, K, L, U_M, eta, ICT, and A. Y denotes real GDP, K denotes capital stock, L denotes employment, U_M denotes aggregate useful exergy, eta denotes final-to-useful conversion efficiency, ICT denotes the ICT investment ratio constructed from OECD national accounts data, and A is an auxiliary variable retained for compatibility with earlier data-construction steps and is not used in the main estimations.

The main notebook, REPLICATION_UsefulWork_OECD_V4.ipynb, reproduces the Mean Group, CCE-MG, AMG, DCCE-MG, LINEX, unit-root, cointegration, robustness, regime-classification, and sensitivity analyses reported in the manuscript and Supplementary Material. The figure notebooks, Figure1_colab_bw.ipynb and Figure2_colab_bw.ipynb, are optional publication-ready plotting scripts that reproduce the black-and-white versions of Figure 1 and Figure 2 from the final manuscript values. These figure scripts are provided for transparency and formatting reproducibility; the main econometric results are generated in the main replication notebook.