Where Chinese Finance Lands, and What It Leaves in Reported Regional Output (Research Square)

Summary

• In the data used, 2,862 projects with combined commitments of about $463 billion (constant 2021 prices) fall inside the regions of 48 countries between 2000 and 2019.p.3
• OOF-like projects number 1,263 but account for 90.0% of allocated value, whereas 1,470 ODA-like projects account for 6.8%.p.7
• In the richest third of regions, 72.0% receive at least one project, against 53.3% in the poorest third.p.12
• For any first project, the average effect on GRP per capita over event years 0 to 4 is −1.01% (95% interval −2.52 to 0.51) in 435 regions and 38 countries.p.16
• With matched controls the any-project estimate becomes +0.18% with the wide caliper and 0.00% with the narrow caliper.p.19
• Russia alone accounts for 24.5% of allocated value, followed by Kazakhstan (8.0%) and Pakistan (7.7%).p.7

Key insights

1. Industry, mining and construction (43.0%), energy (29.5%) and transport (18.0%) absorb over nine-tenths of value.p.7
2. For OOF-like first projects the income and population gradients are 0.87 and 2.20 points, whereas for ODA-like first projects the income gradient is absent (−0.01) and the population gradient is 0.95.p.13
3. The conditional logit gives odds ratios of 1.27 for relative income and 2.58 for population per standard deviation.p.15
4. The point estimate is −1.92% for social projects (p = 0.059), the only coefficient that reaches the 10 percent level.p.16
5. Lead coefficients turn negative at event years −3 (−0.97) and −2 (−0.73), so the post-treatment path continues a pre-treatment drift.p.16
6. Across all specifications and treatment definitions in the robustness table, no confidence interval has an upper bound above 3.85%.p.19
7. Agricultural GRP per capita falls by 3.23% after any first project, but the Holm-adjusted p-values of 0.284 and 0.348 show it does not survive multiple-testing adjustment.p.21