The Cultural Resource Curse: How Trade Dependence Undermines Creative Industries

The Cultural Resource Curse: Replication Package What this is This dataset and code reproduce a global panel analysis testing whether resource dependence (measured as natural resource rents, % of GDP) is associated with weaker cultural/creative export performance, using a country-year panel covering 2000–2023. The analysis controls for trade openness, GDP per capita, internet penetration, secondary school enrollment, population, and institutional quality (legislative constraints on the executive), using two-way fixed-effects panel regression. Robustness checks include Benjamini-Hochberg FDR correction, leave-one-country-out testing, an alternative resource-dependence measure (commodity export composition), static lagged specifications, and a dynamic two-step system-GMM model (Arellano-Bond/Blundell-Bond) with AR(2) and Hansen diagnostics. Files in this repository Culture.ipynb — the complete analysis notebook. Run this to reproduce every result. culture_exports.csv — cultural exports data (UNCTAD/UNESCO source, indicator G.SH.EXP.CLT.ALL) V-Dem-CY-Core-v15.csv — V-Dem Country-Year Core dataset, version 15 (institutional quality data) requirements.txt — exact package versions needed to run the notebook World Bank development indicators (resource rents, trade openness, GDP per capita, internet penetration, secondary enrollment, population, commodity export composition) are not included as files — the notebook pulls them live from the public World Bank API when it runs, so an internet connection is required. How to reproduce this — step by step You do not need anything installed on your own computer. This runs entirely in a free Google Colab notebook in your browser. Step 1: Download the filesFrom this Zenodo record, download all four files: Culture.ipynb, culture_exports.csv, V-Dem-CY-Core-v15.csv, and requirements.txt. Keep the filenames exactly as they are. Step 2: Open the notebook in Colab Go to colab.research.google.com Click File → Upload notebook, and upload Culture.ipynb Step 3: Upload the two data files into your Colab session In the left sidebar of the Colab notebook, click the folder icon Click the upload icon (a page with an up-arrow) near the top of that panel Select culture_exports.csv and V-Dem-CY-Core-v15.csv and upload both. Wait for both to reach 100% before continuing — the V-Dem file is 204 MB and will take a little time. Note: Colab's file storage is temporary. If your session disconnects, you'll need to re-upload these two files before re-running. Step 4: Install dependencies, then restart the runtimeThe notebook's first cell installs the required packages, including a specific version of NumPy (1.26.4) needed for the dynamic-GMM estimation step. After that cell finishes running: Click Runtime in the top menu → Restart session → confirm This restart is required, not optional. Skipping it will cause the GMM portion of the analysis to fail with a NumPy-related error, because Colab keeps the old version of NumPy loaded in memory until the session restarts. Step 5: Run the rest of the notebookAfter restarting, run the remaining cells in order (Runtime → Run all, or run each cell top to bottom individually). The notebook will: Pull World Bank indicators live via API Build the merged country-year panel Estimate the main fixed-effects model and an OECD-interaction model Apply FDR correction and run leave-one-country-out robustness checks Compare the main resource-dependence measure against an alternative Estimate lagged (t-1, t-2) specifications Estimate the dynamic system-GMM model and run leave-one-country-out robustness on it, including individual removal of major resource-exporting countries (Saudi Arabia, Russia, Venezuela, Nigeria, Kuwait, Qatar, Angola, Iraq, Algeria, Kazakhstan) Step 6: Get your resultsAll output tables (model summaries, leave-one-out results, FDR-corrected p-values, the merged panel dataset itself) are saved to an outputs folder inside the Colab session and also printed directly in the notebook as it runs. To download a file, open the folder icon in the sidebar, navigate to outputs, right-click any file, and choose Download. Estimated run time Most of the notebook completes in a few minutes. The GMM leave-one-country-out step is the slowest part — it fits a separate model once per country in the sample (~140+ full model fits) and can take several minutes on its own. Questions For questions about the data, code, or methodology, contact the corresponding author.