#!/usr/bin/env python3 """ Governance Boundaries Test -- Paper 2, Section 6 ================================================== Tests the CES integration boundary predictions: industries with more complementary/relationship-specific inputs should have more vertical integration (higher VA/GO ratio). Sub-tests: A. Contract intensity predicts VA/GO (Nunn, Proposition 6.6) B. Sigma predicts VA/GO negatively (Proposition 6.5) C. Horse race: CI and sigma jointly predict VA/GO D. Nunn cross-country: ECI ~ Rule of Law (K_eff micro-foundation) E. Cross-country with GDP/capita control: ECI ~ Rule of Law | log(GDP/capita) Lean theorem: integration_boundary (CESProofs/Potential/FirmScope.lean) K_eff > c₀ * T ↔ K_eff / T > c₀ (REAL PROOF) Review Notes (2026-03-12): Theory match: PARTIAL — tests monotone correlates of integration boundary, not boundary itself I/O audit: CORRECT — all variables correctly mapped (VA/GO for integration, CI for specificity, sigma for substitutability, Rule of Law for institutional quality) Data-signal alignment: ADEQUATE — BEA VA/GO direct; Nunn CI established (Nunn 2007 QJE); USITC sigma model-dependent; ECI/WGI direct Code quality: CLEAN — minor: Test A should use one-sided p (fixed 2026-03-12) Confidence: HIGH — all 3 subtests CONSISTENT; Spearman confirms OLS in all cases Known limitations: N=25 (Test A) and N=19 (Test B) are small; 3-digit NAICS concordance lossy Migrated from scripts/test_governance_boundaries.py """ import warnings import numpy as np import pandas as pd from scipy import stats from common.config import COMPILED_DIR, RESULTS_DIR from common.plotting import setup_figure, setup_subplots, save_figure from common.stats import verdict, log_result warnings.filterwarnings('ignore') def load_bea_va_go(): """Load BEA VA/GO ratio data.""" try: return pd.read_csv(COMPILED_DIR / 'bea_va_go_2019.csv') except FileNotFoundError: return None def load_nunn(): """Load Nunn (2007) contract intensity data.""" try: return pd.read_csv(COMPILED_DIR / 'nunn_contract_intensity.csv') except FileNotFoundError: return None def load_usitc_sigma(): """Load USITC Ahmad-Riker sigma estimates.""" try: df = pd.read_excel(COMPILED_DIR / 'usitc_sigma.xlsx') df.columns = [c.lower().strip() for c in df.columns] return df except FileNotFoundError: return None def build_concordance(nunn): """Aggregate Nunn to 3-digit and 2-digit NAICS levels.""" nunn = nunn.copy() nunn['naics3'] = nunn['industry_code'].str[:3].astype(str) nunn['naics2'] = nunn['industry_code'].str[:2].astype(str) nunn_3d = nunn.groupby('naics3').agg( frac_lib_diff=('frac_lib_diff', 'mean'), n_industries=('industry_code', 'count'), ).reset_index() nunn_2d = nunn.groupby('naics2').agg( frac_lib_diff=('frac_lib_diff', 'mean'), n_industries=('industry_code', 'count'), ).reset_index() return nunn_3d, nunn_2d def test_contract_integration(bea, nunn): """Test A: Contract intensity predicts vertical integration.""" if bea is None or nunn is None: return log_result('governance_boundaries_ci_va_go', np.nan, np.nan, 'SKIPPED', description="Tests Proposition 6.6: industries with more relationship-specific inputs (higher Nunn contract intensity) have more vertical integration (higher VA/GO ratio). OLS regression of BEA VA/GO on Nunn (2007) fraction of differentiated inputs at 3-digit NAICS level.") if 'frac_lib_diff' not in nunn.columns: return log_result('governance_boundaries_ci_va_go', np.nan, np.nan, 'SKIPPED', description="Tests Proposition 6.6: industries with more relationship-specific inputs (higher Nunn contract intensity) have more vertical integration (higher VA/GO ratio). OLS regression of BEA VA/GO on Nunn (2007) fraction of differentiated inputs at 3-digit NAICS level.") nunn_3d, nunn_2d = build_concordance(nunn) bea = bea.copy() bea['naics_raw'] = bea['industry'].str.extract(r'^(\d+)', expand=False) bea['naics3'] = bea['naics_raw'].str[:3] merged = bea.merge(nunn_3d, on='naics3', how='inner') merged = merged.dropna(subset=['va_go_ratio', 'frac_lib_diff']) merged = merged[(merged['va_go_ratio'] > 0.05) & (merged['va_go_ratio'] < 1.0)] if len(merged) < 5: # Fall back to 2-digit bea['naics2'] = bea['naics_raw'].str[:2] merged = bea.merge(nunn_2d, on='naics2', how='inner') merged = merged.dropna(subset=['va_go_ratio', 'frac_lib_diff']) merged = merged[(merged['va_go_ratio'] > 0.05) & (merged['va_go_ratio'] < 1.0)] if len(merged) < 5: return log_result('governance_boundaries_ci_va_go', np.nan, np.nan, 'SKIPPED', description="Tests Proposition 6.6: industries with more relationship-specific inputs (higher Nunn contract intensity) have more vertical integration (higher VA/GO ratio). OLS regression of BEA VA/GO on Nunn (2007) fraction of differentiated inputs at 3-digit NAICS level.") slope, intercept, r_val, p_val, std_err = stats.linregress( merged['frac_lib_diff'], merged['va_go_ratio']) sign_ok = slope > 0 # more contract-intensive => more integration p_one = p_val / 2 if sign_ok else 1 - p_val / 2 # one-sided: theory predicts positive rho_s, p_s = stats.spearmanr(merged['frac_lib_diff'], merged['va_go_ratio']) verd = verdict(p_one, sign_ok, threshold=0.10) _cache['ci_va_go'] = {'x': merged['frac_lib_diff'].values, 'y': merged['va_go_ratio'].values, 'slope': slope, 'intercept': intercept, 'p': p_one, 'rho_s': rho_s, 'p_s': p_s, 'n': len(merged)} return log_result('governance_boundaries_ci_va_go', slope, p_one, verd, details={'n': len(merged), 'r2': r_val**2, 'spearman_rho': rho_s, 'spearman_p': p_s}, description="Tests Proposition 6.6: industries with more relationship-specific inputs (higher Nunn contract intensity) have more vertical integration (higher VA/GO ratio). OLS of BEA VA/GO on Nunn (2007) fraction of differentiated inputs at 3-digit NAICS. One-sided test (theory predicts positive slope). Prediction: slope > 0.") def test_sigma_integration(bea, usitc): """Test B: Lower sigma => higher VA/GO.""" if bea is None or usitc is None: return log_result('governance_boundaries_sigma_va_go', np.nan, np.nan, 'SKIPPED', description="Tests Proposition 6.5: industries with lower elasticity of substitution sigma have more vertical integration. OLS regression of BEA VA/GO on USITC Ahmad-Riker sigma estimates at 3-digit NAICS; expects negative slope (one-sided test).") if 'sigma1' not in usitc.columns: return log_result('governance_boundaries_sigma_va_go', np.nan, np.nan, 'SKIPPED', description="Tests Proposition 6.5: industries with lower elasticity of substitution sigma have more vertical integration. OLS regression of BEA VA/GO on USITC Ahmad-Riker sigma estimates at 3-digit NAICS; expects negative slope (one-sided test).") usitc = usitc.copy() usitc['naics_str'] = usitc['naics'].astype(str) usitc['naics3'] = usitc['naics_str'].str[:3] usitc_3d = usitc.groupby('naics3').agg(sigma1=('sigma1', 'mean')).reset_index() bea = bea.copy() bea['naics_raw'] = bea['industry'].str.extract(r'^(\d+)', expand=False) bea['naics3'] = bea['naics_raw'].str[:3] merged = bea.merge(usitc_3d, on='naics3', how='inner') merged = merged.dropna(subset=['va_go_ratio', 'sigma1']) merged = merged[(merged['va_go_ratio'] > 0.05) & (merged['va_go_ratio'] < 1.0)] if len(merged) < 5: return log_result('governance_boundaries_sigma_va_go', np.nan, np.nan, 'SKIPPED', description="Tests Proposition 6.5: industries with lower elasticity of substitution sigma have more vertical integration. OLS regression of BEA VA/GO on USITC Ahmad-Riker sigma estimates at 3-digit NAICS; expects negative slope (one-sided test).") slope, intercept, r_val, p_val, std_err = stats.linregress( merged['sigma1'], merged['va_go_ratio']) sign_ok = slope < 0 # lower sigma (more complementary) => more integration p_one = p_val / 2 if sign_ok else 1 - p_val / 2 rho_s, p_s = stats.spearmanr(merged['sigma1'], merged['va_go_ratio']) verd = verdict(p_one, sign_ok, threshold=0.10) _cache['sigma_va_go'] = {'x': merged['sigma1'].values, 'y': merged['va_go_ratio'].values, 'slope': slope, 'intercept': intercept, 'p': p_one, 'rho_s': rho_s, 'p_s': p_s, 'n': len(merged)} return log_result('governance_boundaries_sigma_va_go', slope, p_one, verd, details={'n': len(merged), 'r2': r_val**2, 'spearman_rho': rho_s, 'spearman_p': p_s}, description="Tests Proposition 6.5: industries with lower elasticity of substitution sigma have more vertical integration. OLS of BEA VA/GO on USITC Ahmad-Riker sigma estimates at 3-digit NAICS. One-sided test (theory predicts negative slope). Prediction: slope < 0.") def _prepare_bea(bea): """Add NAICS code columns to BEA data.""" bea = bea.copy() bea['naics_raw'] = bea['industry'].str.extract(r'^(\d+)', expand=False) bea['naics3'] = bea['naics_raw'].str[:3] bea['naics2'] = bea['naics_raw'].str[:2] return bea def test_horse_race(bea, nunn, usitc): """Test C: CI and sigma jointly predict VA/GO.""" skip_desc = ("Tests whether contract intensity and sigma independently predict " "vertical integration. Multiple OLS of BEA VA/GO on Nunn CI + USITC " "sigma at 3-digit NAICS. Both channels should be significant jointly " "(CI positive, sigma negative).") if bea is None or nunn is None or usitc is None: return log_result('governance_boundaries_horse_race', np.nan, np.nan, 'SKIPPED', description=skip_desc) nunn_3d, _ = build_concordance(nunn) usitc2 = usitc.copy() usitc2['naics3'] = usitc2['naics'].astype(str).str[:3] usitc_3d = usitc2.groupby('naics3').agg(sigma1=('sigma1', 'mean')).reset_index() bea2 = _prepare_bea(bea) merged = bea2.merge(nunn_3d, on='naics3', how='inner') merged = merged.merge(usitc_3d, on='naics3', how='inner') merged = merged.dropna(subset=['va_go_ratio', 'frac_lib_diff', 'sigma1']) merged = merged[(merged['va_go_ratio'] > 0.05) & (merged['va_go_ratio'] < 1.0)] if len(merged) < 8: return log_result('governance_boundaries_horse_race', np.nan, np.nan, 'SKIPPED', description=skip_desc) # Multiple regression: VA/GO ~ CI + sigma y = merged['va_go_ratio'].values X = np.column_stack([np.ones(len(merged)), merged['frac_lib_diff'].values, merged['sigma1'].values]) beta, res, rank, sv = np.linalg.lstsq(X, y, rcond=None) y_hat = X @ beta rss = np.sum((y - y_hat) ** 2) tss = np.sum((y - y.mean()) ** 2) r2 = 1 - rss / tss n, k = len(y), 3 r2_adj = 1 - (1 - r2) * (n - 1) / (n - k) mse = rss / (n - k) se = np.sqrt(np.diag(mse * np.linalg.inv(X.T @ X))) # t-tests for CI (beta[1]) and sigma (beta[2]) t_ci = beta[1] / se[1] t_sigma = beta[2] / se[2] from scipy.stats import t as t_dist p_ci = t_dist.sf(abs(t_ci), n - k) # one-sided (positive) p_sigma = t_dist.sf(abs(t_sigma), n - k) # one-sided (negative) sign_ci_ok = beta[1] > 0 sign_sigma_ok = beta[2] < 0 # F-test for joint significance f_stat = ((tss - rss) / (k - 1)) / mse from scipy.stats import f as f_dist p_f = 1 - f_dist.cdf(f_stat, k - 1, n - k) # Verdict: both channels correct sign AND joint F significant both_signs = sign_ci_ok and sign_sigma_ok verd = verdict(p_f, both_signs, threshold=0.10) _cache['horse_race'] = { 'ci_vals': merged['frac_lib_diff'].values, 'sigma_vals': merged['sigma1'].values, 'va_go': merged['va_go_ratio'].values, 'beta_ci': beta[1], 'se_ci': se[1], 'p_ci': p_ci, 'beta_sigma': beta[2], 'se_sigma': se[2], 'p_sigma': p_sigma, 'r2_adj': r2_adj, 'f_stat': f_stat, 'p_f': p_f, 'n': n, } return log_result('governance_boundaries_horse_race', f_stat, p_f, verd, details={'n': n, 'r2_adj': r2_adj, 'beta_ci': beta[1], 'p_ci_onesided': p_ci, 'beta_sigma': beta[2], 'p_sigma_onesided': p_sigma}, description=("Tests whether contract intensity and sigma independently " "predict vertical integration. Multiple OLS: VA/GO ~ CI + sigma " f"at 3-digit NAICS (N={n}). F-test for joint significance. " "Prediction: CI positive, sigma negative, joint F significant.")) def test_nunn_cross_country(): """Test D: ECI ~ Rule of Law (K_eff micro-foundation).""" try: eci = pd.read_csv(COMPILED_DIR / 'eci_rankings.csv') wgi = pd.read_csv(COMPILED_DIR / 'wgi_2019.csv') except FileNotFoundError: return log_result('governance_boundaries_nunn_cross_country', np.nan, np.nan, 'SKIPPED', description="Tests the K_eff micro-foundation across countries: Economic Complexity Index (ECI) should increase with Rule of Law (WGI). OLS regression of ECI on WGI Rule of Law across countries; positive slope means better institutions support more complex production.") if 'rule_of_law' not in wgi.columns or 'eci_hs92' not in eci.columns: return log_result('governance_boundaries_nunn_cross_country', np.nan, np.nan, 'SKIPPED', description="Tests the K_eff micro-foundation across countries: Economic Complexity Index (ECI) should increase with Rule of Law (WGI). OLS regression of ECI on WGI Rule of Law across countries; positive slope means better institutions support more complex production.") df = eci.merge(wgi, on='iso3', how='inner') df = df.dropna(subset=['eci_hs92', 'rule_of_law']) if len(df) < 20: return log_result('governance_boundaries_nunn_cross_country', np.nan, np.nan, 'SKIPPED', description="Tests the K_eff micro-foundation across countries: Economic Complexity Index (ECI) should increase with Rule of Law (WGI). OLS regression of ECI on WGI Rule of Law across countries; positive slope means better institutions support more complex production.") slope, intercept, r_val, p_val, std_err = stats.linregress( df['rule_of_law'], df['eci_hs92']) sign_ok = slope > 0 # better institutions => more complex exports rho_s, p_s = stats.spearmanr(df['rule_of_law'], df['eci_hs92']) verd = verdict(p_val, sign_ok, threshold=0.10) _cache['nunn_cross_country'] = {'x': df['rule_of_law'].values, 'y': df['eci_hs92'].values, 'slope': slope, 'intercept': intercept, 'p': p_val, 'rho_s': rho_s, 'p_s': p_s, 'n': len(df)} return log_result('governance_boundaries_nunn_cross_country', slope, p_val, verd, details={'n': len(df), 'r2': r_val**2, 'spearman_rho': rho_s, 'spearman_p': p_s}, description="Tests the K_eff micro-foundation: Economic Complexity Index (ECI) should increase with Rule of Law (WGI). OLS of ECI on WGI Rule of Law across 144 countries. Better institutions reduce information friction T, increasing K_eff and enabling more complex production. Prediction: slope > 0.") def test_nunn_cross_country_controlled(): """Test E: ECI ~ Rule of Law controlling for GDP/capita.""" skip_desc = ("Tests whether Rule of Law predicts ECI beyond the income gradient. " "Multiple OLS: ECI ~ Rule of Law + log(GDP/capita). If Rule of Law " "remains significant after controlling for income, the K_eff " "micro-foundation holds beyond development level.") try: eci = pd.read_csv(COMPILED_DIR / 'eci_rankings.csv') wgi = pd.read_csv(COMPILED_DIR / 'wgi_2019.csv') gdp = pd.read_csv(COMPILED_DIR / 'gdp_pc_2019.csv') except FileNotFoundError: return log_result('governance_boundaries_nunn_controlled', np.nan, np.nan, 'SKIPPED', description=skip_desc) df = eci.merge(wgi, on='iso3', how='inner').merge(gdp, on='iso3', how='inner') df = df.dropna(subset=['eci_hs92', 'rule_of_law', 'gdp_pc_2019']) df = df[df['gdp_pc_2019'] > 0] df['log_gdppc'] = np.log(df['gdp_pc_2019']) if len(df) < 30: return log_result('governance_boundaries_nunn_controlled', np.nan, np.nan, 'SKIPPED', description=skip_desc) # Multiple regression: ECI ~ Rule of Law + log(GDP/capita) y = df['eci_hs92'].values X = np.column_stack([np.ones(len(df)), df['rule_of_law'].values, df['log_gdppc'].values]) beta, res, rank, sv = np.linalg.lstsq(X, y, rcond=None) y_hat = X @ beta rss = np.sum((y - y_hat) ** 2) tss = np.sum((y - y.mean()) ** 2) r2 = 1 - rss / tss n, k = len(y), 3 r2_adj = 1 - (1 - r2) * (n - 1) / (n - k) mse = rss / (n - k) se = np.sqrt(np.diag(mse * np.linalg.inv(X.T @ X))) # t-test for Rule of Law coefficient (beta[1]) from scipy.stats import t as t_dist t_rol = beta[1] / se[1] p_rol = 2 * t_dist.sf(abs(t_rol), n - k) # two-sided sign_ok = beta[1] > 0 # Also get the bivariate for comparison slope_biv, _, r_biv, p_biv, _ = stats.linregress(df['rule_of_law'], df['eci_hs92']) verd = verdict(p_rol, sign_ok, threshold=0.10) _cache['nunn_controlled'] = { 'x_rol': df['rule_of_law'].values, 'x_gdp': df['log_gdppc'].values, 'y': df['eci_hs92'].values, 'beta_rol': beta[1], 'se_rol': se[1], 'p_rol': p_rol, 'beta_gdp': beta[2], 'se_gdp': se[2], 'r2_adj': r2_adj, 'n': n, 'slope_bivariate': slope_biv, 'p_bivariate': p_biv, } return log_result('governance_boundaries_nunn_controlled', beta[1], p_rol, verd, details={'n': n, 'r2_adj': r2_adj, 'beta_rol': beta[1], 'p_rol': p_rol, 'beta_gdppc': beta[2], 'se_gdppc': se[2], 'bivariate_slope': slope_biv, 'bivariate_p': p_biv}, description=("Tests whether Rule of Law predicts ECI beyond income. " f"Multiple OLS: ECI ~ Rule of Law + log(GDP/capita) (N={n}). " "If Rule of Law remains significant after controlling for " "development level, the K_eff channel operates independently. " "Prediction: Rule of Law coefficient > 0.")) _cache = {} def main(): results = [] bea = load_bea_va_go() nunn = load_nunn() usitc = load_usitc_sigma() results.append(test_contract_integration(bea, nunn)) results.append(test_sigma_integration(bea, usitc)) results.append(test_horse_race(bea, nunn, usitc)) results.append(test_nunn_cross_country()) results.append(test_nunn_cross_country_controlled()) _cache['results'] = results return results def figures(): """Generate figures for each subtest.""" if not _cache: main() paths = [] # --- Figure for ci_va_go --- data = _cache.get('ci_va_go') if data is not None: fig, ax = setup_figure() ax.scatter(data['x'], data['y'], alpha=0.6, s=50) x_line = np.linspace(data['x'].min(), data['x'].max(), 100) ax.plot(x_line, data['intercept'] + data['slope'] * x_line, 'r-', linewidth=2, label=f"OLS slope={data['slope']:.3f}, p={data['p']:.3f}") ax.set_xlabel('Nunn Contract Intensity (frac differentiated)') ax.set_ylabel('BEA VA/GO Ratio') ax.set_title(f"Contract Intensity vs Vertical Integration (N={data['n']})") ax.annotate(f"Spearman r={data['rho_s']:.3f}, p={data['p_s']:.3f}", xy=(0.05, 0.95), xycoords='axes fraction', va='top', fontsize=9) ax.legend(loc='lower right', fontsize=9) paths.extend(save_figure(fig, 'governance_boundaries_ci_va_go', subfolder='applications')) # --- Figure for sigma_va_go --- data = _cache.get('sigma_va_go') if data is not None: fig, ax = setup_figure() ax.scatter(data['x'], data['y'], alpha=0.6, s=50) x_line = np.linspace(data['x'].min(), data['x'].max(), 100) ax.plot(x_line, data['intercept'] + data['slope'] * x_line, 'r-', linewidth=2, label=f"OLS slope={data['slope']:.3f}, p={data['p']:.4f}") ax.set_xlabel('USITC Sigma (elasticity of substitution)') ax.set_ylabel('BEA VA/GO Ratio') ax.set_title(f"Sigma vs Vertical Integration (N={data['n']})") ax.annotate(f"Spearman r={data['rho_s']:.3f}, p={data['p_s']:.4f}", xy=(0.05, 0.95), xycoords='axes fraction', va='top', fontsize=9) ax.legend(loc='upper right', fontsize=9) paths.extend(save_figure(fig, 'governance_boundaries_sigma_va_go', subfolder='applications')) # --- Figure for horse_race --- data = _cache.get('horse_race') if data is not None: fig, axes = setup_subplots(nrows=1, ncols=2, width=12, height=5) # Left: partial effect of CI axes[0].scatter(data['ci_vals'], data['va_go'], alpha=0.6, s=50) axes[0].set_xlabel('Nunn Contract Intensity') axes[0].set_ylabel('BEA VA/GO Ratio') axes[0].set_title(f"CI channel: beta={data['beta_ci']:.3f}, p={data['p_ci']:.3f}") # Right: partial effect of sigma axes[1].scatter(data['sigma_vals'], data['va_go'], alpha=0.6, s=50) axes[1].set_xlabel('USITC Sigma') axes[1].set_ylabel('BEA VA/GO Ratio') axes[1].set_title(f"Sigma channel: beta={data['beta_sigma']:.3f}, p={data['p_sigma']:.3f}") fig.suptitle(f"Horse Race: CI + Sigma (N={data['n']}, adj-R²={data['r2_adj']:.3f}, F p={data['p_f']:.3f})", fontsize=11, y=1.02) fig.tight_layout() paths.extend(save_figure(fig, 'governance_boundaries_horse_race', subfolder='applications')) # --- Figure for nunn_cross_country --- data = _cache.get('nunn_cross_country') if data is not None: fig, ax = setup_figure() ax.scatter(data['x'], data['y'], alpha=0.4, s=30) x_line = np.linspace(data['x'].min(), data['x'].max(), 100) ax.plot(x_line, data['intercept'] + data['slope'] * x_line, 'r-', linewidth=2, label=f"OLS slope={data['slope']:.3f}, p={data['p']:.1e}") ax.set_xlabel('WGI Rule of Law') ax.set_ylabel('Economic Complexity Index (ECI)') ax.set_title(f"Rule of Law vs Economic Complexity (N={data['n']})") ax.annotate(f"Spearman r={data['rho_s']:.3f}, p={data['p_s']:.1e}", xy=(0.05, 0.95), xycoords='axes fraction', va='top', fontsize=9) ax.legend(loc='lower right', fontsize=9) paths.extend(save_figure(fig, 'governance_boundaries_nunn_cross_country', subfolder='applications')) # --- Figure for nunn_controlled --- data = _cache.get('nunn_controlled') if data is not None: fig, axes = setup_subplots(nrows=1, ncols=2, width=12, height=5) # Left: bivariate (no control) axes[0].scatter(data['x_rol'], data['y'], alpha=0.4, s=30) x_line = np.linspace(data['x_rol'].min(), data['x_rol'].max(), 100) axes[0].plot(x_line, np.polyval(np.polyfit(data['x_rol'], data['y'], 1), x_line), 'r-', linewidth=2) axes[0].set_xlabel('WGI Rule of Law') axes[0].set_ylabel('ECI') axes[0].set_title(f"Bivariate: slope={data['slope_bivariate']:.3f}, p={data['p_bivariate']:.1e}") # Right: partial residual plot (ECI residualized on GDP, plotted vs RoL) y_resid = data['y'] - np.polyval(np.polyfit(data['x_gdp'], data['y'], 1), data['x_gdp']) axes[1].scatter(data['x_rol'], y_resid, alpha=0.4, s=30) x_line = np.linspace(data['x_rol'].min(), data['x_rol'].max(), 100) axes[1].plot(x_line, np.polyval(np.polyfit(data['x_rol'], y_resid, 1), x_line), 'r-', linewidth=2) axes[1].set_xlabel('WGI Rule of Law') axes[1].set_ylabel('ECI (residualized on log GDP/capita)') axes[1].set_title(f"Controlled: beta_RoL={data['beta_rol']:.3f}, p={data['p_rol']:.3f}") fig.suptitle(f"Rule of Law vs ECI: With and Without Income Control (N={data['n']})", fontsize=11, y=1.02) fig.tight_layout() paths.extend(save_figure(fig, 'governance_boundaries_nunn_controlled', subfolder='applications')) return paths if __name__ == '__main__': results = main() for r in results: print(f"{r['name']}: {r['verdict']} " f"(stat={r['statistic']}, p={r['p_value']})")

n	144
r2	0.6367
p value	4.999e-33
statistic	0.8383

Governance Boundaries Nunn Cross Country

Description

Figures

Statistics

Script

▸ Source Code

Reproduction Steps