Development Goals, Emissions Costs? Climate and Development Finance vis-à-vis Local CO₂ Emissions

Can development finance simultaneously deliver growth and reduce emissions at the subnational level, or does the development–carbon trade-off persist where investments actually land? This paper addresses this question by combining a formal model of local energy substitution with a large-scale quasi-experimental evaluation of geocoded aid. On the theory side, I develop a model of a local economy with clean–dirty energy substitution and a public finance transmission mechanism. The model yields propositions on the conditions under which non-climate aid raises emissions, climate finance achieves abatement, and win–win portfolios exist. On the empirical side, I construct a global ADM2-year panel linking geocoded development projects to local CO₂ emissions and night-time lights. Climate relevance is measured at the project level using ClimateFinanceBERT, a domain-specific classifier that separates mitigation, adaptation, and non-climate flows consistently across donors and years. I estimate causal effects using staggered not-yet-treated doubly-robust difference-in-differences, exploiting the quasi-random timing of first project arrival across subnational units.