Replication package for "Green Foreign Direct Investment is flowing far beyond renewables: a taxonomy-guided LLM analysis"

This release accompanies the submission of "Green Foreign Direct Investment is flowing far beyond renewables: a taxonomy-guided LLM analysis". The replication package provides a full end-to-end pipeline for taxonomy-guided classification of greenfield FDI projects using large language models. It includes all code, prompt templates, and curated EU Taxonomy sector mappings required to reproduce the classification, evaluation exercises, and descriptive analyses reported in the manuscript and Supplementary Information. To request access to the replication package, please email j.alvarez-vilanova@lse.ac.uk. Contents Modular Python pipeline for data processing, LLM classification, and master dataset construction Strict and basic EU Taxonomy sector mappings (February 2025 extract) Full prompt templates for both classification stages Evaluation scripts reproducing accuracy, validity, and reproducibility analyses Scripts generating all figures and tables in the manuscript and SI Data note: Project-level FDI data from Orbis-Crossborder cannot be shared publicly due to licensing restrictions. Users with institutional access can reproduce the dataset using the export instructions provided.