Earthquakes and the Wealth of Nations

This repository contains the analysis and code for the paper "Earthquakes and the Wealth of Nations: The cases of Chile and New Zealand" (Díaz, Paniagua, Larroulet).

Reproducibility

Regenerating all figures

All figures used by main.tex are produced by Python code and saved directly to article_assets/. To regenerate all figures:

python run_analysis.py

This script:

1. Runs src/create_maps.py (Maule and Canterbury maps)
2. Runs src/nz_outcome_extensions.py (GDP / population decomposition)
3. Runs src/sdid_bias_corrected_analysis.py (SDID + penalized SCM robustness)
4. Runs src/uniform_confidence_analysis.py (uniform confidence sets + sensitivity checks)
5. Runs src/treatment_timing_sensitivity.py (2010/2011 timing diagnostics + placebo ranks)
6. Runs src/nighttime_lights_validation.py (independent night-time lights validation)
7. Runs src/main_scm_figures.py (main SCM paths, gaps, placebos, jackknife for NZ and Chile)
8. Runs src/sectoral_appendix_analysis.py (parallel Chile/NZ sectoral SCM appendix outputs)
9. Runs src/predictor_weight_sensitivity.py (predictor-weight tuning + harmonized predictor-set sensitivity)

Note: the night-time lights validation step downloads yearly global rasters on first run and caches them under data/ntl/rasters/ (large, ignored by git). Subsequent runs reuse the local cache.

Figure output

Figures are written to article_assets/, which is the canonical location expected by main.tex:

• Maps: Maule_map.png, Canterbury_map.png
• GDP paths: maule_gdp_paths.png, nz_gdp_paths.png
• Gaps: maule_gap.png, nz_gap.png
• Placebos: maule_placebos.png, nz_placebos.png
• Uniform confidence sets: scm_uniform_confidence_sets.png, chile_uniform_threshold_sensitivity.png, nz_uniform_threshold_sensitivity.png
• Sectoral: nz_scm_Construction.png, nz_scm_Other_Sectors.png
• Sectoral appendix (new): chile_scm_Construction.png, chile_scm_Other_Sectors.png, sectoral_inference_summary.csv, sectoral_crowding_out_summary.csv, sectoral_grouping_sensitivity.csv, sectoral_appendix_series.xlsx
• Predictor-weight / predictor-symmetry sensitivity: predictor_spec_sensitivity.csv
• Jackknife: chile_jacknife.png, nz_jacknife.png
• SDID / bias-corrected robustness: sdid_bias_corrected_summary.csv, sdid_bias_corrected_gaps.png
• Uniform confidence tables: scm_uniform_confidence_sets.csv, scm_uniform_confidence_summary.csv
• Timing sensitivity appendix outputs: timing_sensitivity_summary.csv, timing_sensitivity_gap_paths.png, timing_sensitivity_rmspe_ratios.png
• Night-time lights validation: ntl_validation_paths_gaps.png, ntl_sensor_processing_robustness.png, ntl_spatial_sensitivity.png, ntl_validation_summary.csv, ntl_scm_summary.csv, ntl_scm_gaps.csv

Requirements

See requirements.txt. Key dependencies: pysyncon, pandas, matplotlib, geopandas.