Reflections eight years on from the first declaration of climate emergency: The role of LDA topic modelling combined with qualitative policy analysis in detecting a frame of "climate emergency" in real-world policy

Abstract

This paper considers the merits of combining LDA topic modeling as a technique in Natural Language Processing [NLP] and policy analysis to contribute to methods for systematically analysing the rapidly evolving climate policy landscape. The novelty of the use of topic modelling within our methods contributes to a growing literature on using NLP to analyse policy changes. In our case study, we consider the policy frame of “climate emergency”. Eight years after the first declaration of climate emergency and now with the movement slowing, it is timely to reflect on the presence (or not) of the climate emergency policy mode. To do so, we undertake a text analysis of local government climate strategy documents of 70 local governments in Australia that declared a climate emergency from 2016 to the end of 2022. We aim to ascertain whether the framing of "climate emergency" can be detected in real-world policy utilising a mixed-methods approach of qualitative and quantitative methods, including the use of LDA topic modelling and qualitative policy analysis. We conclude that topic modelling techniques such as LDA contribute to the identification and analysis of the evolving framing of “climate emergency” in local governments’ policies. LDA Topic modelling complements traditional qualitative policy analysis by introducing efficient and replicable methods for comprehensive examination of policy documents. In addition, for the first time at scale, we can assess the impact of the Climate Emergency Declaration movement across local governments in Australia, revealing the presence of all key attributes of the climate emergency mode.