Within the European Green Deal (EGD), how do mentions of different types of organizations (e.g. international organizations, decentralized agencies, and financial bodies) correlate with underlying expectations about their performance in climate governance (e.g. sentiment, commitment, specificity, net zero reduction, risk, and strategy)? Our research examines this question by means of Natural Language Processing (NLP) analyses of the 14 legal EGD documents, split into 804 sections. We also analyze the 37 proposal texts and the European Climate Law, as benchmarks. Climate relation of each text section is based on ClimateBert, a specialized large language model for climate-specific texts. Organizations were identified through a customized Named Entity Recognition model. We recorded official organization types from the EU and UN and cross-referenced them with the organizations referenced in the EGD. This methodological innovation ensures precision in identifying actors and provides a framework for future research on larger corpus, such as broader analyses of media. Our work expands the suite of text-as-data tools to analyze environmental policy discourse. Our method is simple: we conduct a t-test to examine if there is a significant difference in climate policy relations when an organization of interest (e.g. the United Nations) is present in a section of the EGD. We then verify if such correlations are sustained within the proposals and European Climate Law. We discuss how different types of actors at a geopolitical level are correlated with expectations of climate policy-making. We highlight how distinct types of organizations shape multinational climate policy.

This research was funded by the German Research Foundation (DFG) , project “The Emergence of Global Administrative Spaces in Transnational Climate Policy,” Reference No. KO 4997/10-1.

global administrative space,European Green Deal,environmental policy,climate policy,natural language processing,named entity recognition,text-as-data,machine learning