War and Public Spending: Conflict, Efficiency, and State Capacity in Developing Countries

Description

This package contains the complete Stata code and data to reproduce every table and figure in the paper and its appendices. Using entropy balancing on 124 developing countries over 1990-2017, the paper estimates the effect of armed conflict on public spending efficiency, and identifies institutional erosion as a central transmission channel. A single master do-file builds the dataset from the raw sources and produces all results, and the README maps each exhibit to the code that generates it. Most inputs are public (UCDP, World Bank WDI and WGI, KOF, Penn World Table, DPI, Polity5, and IMF). Two sources are restricted and cannot be redistributed: the International Country Risk Guide (PRS Group), and sectoral and alternative efficiency measures obtained from the authors of Apeti et al. (2024). For each, a synthetic dataset is provided that preserves the variable names, means, standard deviations, and number of observations, plus instructions for obtaining the originals. With the real restricted data in place, the code reproduces the published results exactly; with the synthetic data, it runs from start to finish but the numbers differ. Software: Stata 18 and 19.

Steps to reproduce

1. Unzip the package and open code/Replication_WarAndPublicSpending.do. 2. Set the single global "root" at the top of the do-file to the unzipped folder. 3. Install the Stata packages listed in the README (SSC). 4. Run the do-file from top to bottom. It builds the analysis dataset and writes all tables and figures to output/. 5. For Figure 7 (correlation matrix), run: python code/make_heatmap_correlation.py (Python 3 with pandas, numpy, matplotlib). 4. To reproduce the published numbers, obtain the two restricted sources (ICRG from the PRS Group; sectoral and alternative efficiency measures from the authors of Apeti et al. 2024) and place them in data/raw/. Otherwise keep the provided synthetic files to run the full pipeline with non-identical numbers. 5. Run the do-file from top to bottom. It builds the analysis dataset and writes every table to output/tables/ and every figure to output/figures/. 6. See README Section 6 for the exhibit-to-script mapping.

Institutions

• Université Clermont Auvergne

  Rhône-Alpes, Clermont-Ferrand

Categories

Licence