Energy Intensity, Carbon Intensity, and the Producer–Consumer Divide: Cross-Country Panel Data and Replication Code

A codebook sheet documents each variable's definition, source (World Bank WDI, Penn World Table, IMF WEO, Global Carbon Project, Berkeley Earth/Our World in Data), and role in the analysis. Notable findings: energy-exporting economies draw a partially offsetting productivity return to energy intensity (Petro x ln ENE = +0.143, p<0.05) relative to a robustly negative pooled baseline, while import-dependent economies show no such offset.

Conversely, import-dependent economies realize a significant positive productivity return to carbon intensity (HID x ln EMI = +0.293, p<0.01) that is absent for exporters. These asymmetries survive alternative Petro/HID thresholds, a continuous energy-trade-position specification, and VIF diagnostics, though a dynamic-panel GMM check does not confirm the same interactions once the persistence of GDP itself is absorbed (a limitation discussed in the paper).

How to interpret and reuse: analysis_code.py reproduces every table in the manuscript (Tables 2-6, including the VIF appendix) directly from analysis_data.xlsx via two-way fixed-effects PanelOLS and, for the dynamic-panel check, two-step difference/system GMM (requires "linearmodels" and "pydynpd"). Because each table applies its own listwise deletion, the full unfiltered panel is provided rather than one pre-filtered sample, so every table's exact estimation sample (e.g. N=2,591 for the main specification) can be reconstructed from the code.

See the "readme" sheet in analysis_data.xlsx and the module docstring in analysis_code.py for full documentation.